[{"data":1,"prerenderedAt":793},["ShallowReactive",2],{"/en-us/blog/how-we-spent-two-weeks-hunting-an-nfs-bug":3,"navigation-en-us":40,"banner-en-us":439,"footer-en-us":449,"blog-post-authors-en-us-Stan Hu":691,"blog-related-posts-en-us-how-we-spent-two-weeks-hunting-an-nfs-bug":705,"assessment-promotions-en-us":744,"next-steps-en-us":783},{"id":4,"title":5,"authorSlugs":6,"body":8,"categorySlug":9,"config":10,"content":14,"description":8,"extension":27,"isFeatured":12,"meta":28,"navigation":29,"path":30,"publishedDate":20,"seo":31,"stem":35,"tagSlugs":36,"__hash__":39},"blogPosts/en-us/blog/how-we-spent-two-weeks-hunting-an-nfs-bug.yml","How We Spent Two Weeks Hunting An Nfs Bug",[7],"stan-hu",null,"engineering",{"slug":11,"featured":12,"template":13},"how-we-spent-two-weeks-hunting-an-nfs-bug",false,"BlogPost",{"title":15,"description":16,"authors":17,"heroImage":19,"date":20,"body":21,"category":9,"tags":22},"How we spent two weeks hunting an NFS bug in the Linux kernel","Here's an in-depth recap of debugging a GitLab issue that culminated in a patch for the Linux kernel.",[18],"Stan Hu","https://res.cloudinary.com/about-gitlab-com/image/upload/v1749672173/Blog/Hero%20Images/nfs-bug-hunt-detective.jpg","2018-11-14","UPDATE 2019-08-06: This bug has now been resolved in the following\ndistributions:\n\n* [Red Hat Enterprise Linux 7](https://access.redhat.com/errata/RHSA-2019:2029)\n* [Ubuntu](https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1802585)\n* Linux mainline: Backported to [4.14-stable](https://lkml.org/lkml/2019/8/2/562) and [4.19-stable](https://lkml.org/lkml/2019/8/2/639)\n\nOn Sep. 14, the GitLab support team escalated a critical\nproblem encountered by one of our customers: GitLab would run fine for a\nwhile, but after some time users encountered errors. When attempting to\nclone certain repositories via Git, users would see an opaque `Stale\nfile error` message. The error message persisted for a long time,\nblocking employees from being able to work, unless a system\nadministrator intervened manually by running `ls` in the directory\nitself.\n\nThus launched an investigation into the inner workings of Git and the\nNetwork File System (NFS). The investigation uncovered a bug with the\nLinux v4.0 NFS client and culiminated with a [kernel patch that was written by\nTrond Myklebust](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=be189f7e7f03de35887e5a85ddcf39b91b5d7fc1)\nand [merged in the latest mainline Linux kernel](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=c7a2c49ea6c9eebbe44ff2c08b663b2905ee2c13)\non Oct. 26.\n\nThis post describes the journey of investigating the issue and\ndetails the thought process and tools by which we tracked down the\nbug. It was inspired by the fine detective work in [How I spent two\nweeks hunting a memory leak in Ruby](http://www.be9.io/2015/09/21/memory-leak/)\nby Oleg Dashevskii.\n\nMore importantly, this experience exemplifies how open source software\ndebugging has become a team sport that involves expertise across\nmultiple people, companies, and locations. The GitLab motto \"[everyone can\ncontribute](https://handbook.gitlab.com/handbook/company/mission/#mission)\" applies not only to GitLab itself, but also to other open\nsource projects, such as the Linux kernel.\n\n## Reproducing the bug\n\nWhile we have run NFS on GitLab.com for many years, we have stopped\nusing it to access repository data across our application\nmachines. Instead, we have [abstracted all Git calls to\nGitaly](/blog/the-road-to-gitaly-1-0/).\nStill, NFS remains a supported configuration for our customers who\nmanage their own installation of GitLab, but we had never seen the exact\nproblem described by the customer before.\n\n[Our customer gave us a few important clues](https://gitlab.com/gitlab-org/gitlab-ce/issues/51437):\n\n1. The full error message read, `fatal: Couldn't read ./packed-refs: Stale file handle`.\n2. The error seemed to start when they started a manual Git garbage\ncollection run via `git gc`.\n3. The error would go away if a system administrator ran `ls` in the\ndirectory.\n4. The error also would go away after `git gc` process ended.\n\nThe first two items seemed obviously related. When you push to a branch\nin Git, Git creates a loose reference, a fancy name for a file that\npoints your branch name to the commit. For example, a push to `master`\nwill create a file called `refs/heads/master` in the repository:\n\n```bash\n$ cat refs/heads/master\n2e33a554576d06d9e71bfd6814ee9ba3a7838963\n```\n\n`git gc` has several jobs, but one of them is to collect these loose\nreferences (refs) and bundle them up into a single file called\n`packed-refs`. This makes things a bit faster by eliminating the need to\nread lots of little files in favor of reading one large one. For\nexample, after running `git gc`, an example `packed-refs` might look\nlike:\n\n```text\n# pack-refs with: peeled fully-peeled sorted\n564c3424d6f9175cf5f2d522e10d20d781511bf1 refs/heads/10-8-stable\nedb037cbc85225261e8ede5455be4aad771ba3bb refs/heads/11-0-stable\n94b9323033693af247128c8648023fe5b53e80f9 refs/heads/11-1-stable\n2e33a554576d06d9e71bfd6814ee9ba3a7838963 refs/heads/master\n```\n\nHow exactly is this `packed-refs` file created? To answer that, we ran\n`strace git gc` with a loose ref present. Here are the pertinent lines\nfrom that:\n\n```text\n28705 open(\"/tmp/libgit2/.git/packed-refs.lock\", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666) = 3\n28705 open(\".git/packed-refs\", O_RDONLY) = 3\n28705 open(\"/tmp/libgit2/.git/packed-refs.new\", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666) = 4\n28705 rename(\"/tmp/libgit2/.git/packed-refs.new\", \"/tmp/libgit2/.git/packed-refs\") = 0\n28705 unlink(\"/tmp/libgit2/.git/packed-refs.lock\") = 0\n```\n\nThe system calls showed that `git gc` did the following:\n\n1. Open `packed-refs.lock`. This tells other processes that `packed-refs` is locked and cannot be changed.\n1. Open `packed-refs.new`.\n1. Write loose refs to `packed-refs.new`.\n1. Rename `packed-refs.new` to `packed-refs`.\n1. Remove `packed-refs.lock`.\n1. Remove loose refs.\n\nThe fourth step is the key here: the rename where Git puts `packed-refs`\ninto action. In addition to collecting loose refs, `git gc` also\nperforms a more expensive task of scanning for unused objects and\nremoving them. This task can take over an hour for large\nrepositories.\n\nThat made us wonder: for a large repository, does `git gc` keep the file\nopen while it's running this sweep? Looking at the `strace` logs and\nprobing the process with `lsof`, we found that it did the following:\n\n![Git Garbage Collection](https://about.gitlab.com/images/blogimages/nfs-debug/git-gc-diagram.svg)\n\nNotice that `packed-refs` is closed only at the end, after the potentially\nlong `Garbage collect objects` step takes place.\n\nThat made us wonder: how does NFS behave when one node has `packed-refs`\nopen while another renames over that file?\n\nTo experiment, we asked the customer to run the following experiment on\ntwo different machines (Alice and Bob):\n\n1. On the shared NFS volume, create two files: `test1.txt` and\n`test2.txt` with different contents to make it easy to distinguish them:\n\n    ```bash\n    alice $ echo \"1 - Old file\" > /path/to/nfs/test1.txt\n    alice $ echo \"2 - New file\" > /path/to/nfs/test2.txt\n    ```\n\n2. On machine Alice, keep a file open to `test1.txt`:\n\n    ```bash\n     alice $ irb\n     irb(main):001:0> File.open('/path/to/nfs/test1.txt')\n    ```\n\n3. On machine Alice, show the contents of `test1.txt` continuously:\n\n    ```bash\n    alice $ while true; do cat test1.txt; done\n    ```\n\n4. Then on machine Bob, run:\n\n    ```bash\n    bob $ mv -f test2.txt test1.txt\n    ```\n\nThis last step emulates what `git gc` does with `packed-refs` by\noverwriting the existing file.\n\nOn the customer's machine, the result looked something like:\n\n```text\n1 - Old file\n1 - Old file\n1 - Old file\ncat: test1.txt: Stale file handle\n```\n\nBingo! We seemed to reproduce the problem in a controlled way. However,\nthe same experiment using a Linux NFS server did not have this\nproblem. The result was what you would expect: the new contents were\npicked up after the rename:\n\n```text\n1 - Old file\n1 - Old file\n1 - Old file\n2 - New file  \u003C--- RENAME HAPPENED\n2 - New file\n2 - New file\n```\n\nWhy the difference in behavior? It turns out that the customer was using\nan [Isilon NFS\nappliance](https://www.dellemc.com/en-us/storage/isilon/index.htm) that\nonly supported NFS v4.0. By switching the mount parameters to v4.0 via\nthe `vers=4.0` parameter in `/etc/fstab`, the test revealed a different\nresult with the Linux NFS server:\n\n```text\n1 - Old file\n1 - Old file\n1 - Old file\n1 - Old file \u003C--- RENAME HAPPENED\n1 - Old file\n1 - Old file\n```\n\nInstead of a `Stale file handle`, the Linux NFS v4.0 server showed stale\n*contents*. It turns out this difference in behavior can be explained by\nthe NFS spec. From [RFC\n3010](https://tools.ietf.org/html/rfc3010#page-153):\n\n> A filehandle may or may not become stale or expire on a rename.\n> However, server implementors are strongly encouraged to attempt to keep\n> file handles from becoming stale or expiring in this fashion.\n\nIn other words, NFS servers can choose how to behave if a file is\nrenamed; it's perfectly valid for any NFS server to return a `Stale file\nerror` when that happens. We surmised that even though the results were\ndifferent, the problem was likely related to the same issue. We\nsuspected some cache validation issue because running `ls` in the\ndirectory would \"clear\" the error. Now that we had a reproducible test\ncase, we asked the experts: the Linux NFS maintainers.\n\n## False path: NFS server delegations\n\nWith a clear set of reproduction steps, I [sent an email to the Linux\nNFS mailing list](https://marc.info/?l=linux-nfs&m=153721785231614&w=2)\ndescribing what we had found. Over the week, I went back and forth with\nBruce Fields, the Linux NFS server maintainer, who suggested this was a\nNFS bug and that it would be useful to look at the network traffic. He\nthought there might be an issue with NFS server delegations.\n\n### What is an NFS server delegation?\n\nIn a nutshell, NFS v4 introduced server delegations as a way to speed up file access. A server can\ndelegate read or write access to a client so that the client doesn't\nhave to keep asking the server whether that file has changed by another\nclient. In simpler terms, a write delegation is akin to someone lending\nyou a notebook and saying, \"Go ahead and write in here, and I'll take it\nback when I'm ready.\" Instead of having to ask to borrow the notebook\nevery time you want to write a new paragraph, you have free rein until\nthe owner reclaims the notebook. In NFS terms, this reclamation process\nis called a delegation recall.\n\nIndeed, a bug in the NFS delegation recall might explain the `Stale file\nhandle` problem. Remember that in the earlier experiment, Alice had\nan open file to `test1.txt` when it was replaced by `test2.txt` later.\nIt's possible that the server failed to recall the delegation on\n`test1.txt`, resulting in an incorrect state. To check whether this was\nan issue, we turned to `tcpdump` to capture NFS traffic and used\nWireshark to visualize it.\n\n[Wireshark](https://www.wireshark.org/) is a wonderful open source tool\nfor analyzing network traffic, and it's especially good for viewing NFS\nin action. We captured a trace using the following command on the NFS server:\n\n```text\ntcpdump -s 0 -w /tmp/nfs.pcap port 2049\n```\n\nThis command captures all NFS traffic, which typically is on TCP port 2049.\nBecause our experiment worked properly with NFS v4.1 but did not\n with NFS v4.0, we could compare and contrast how NFS behaved\nin a non-working and a working case. With Wireshark, we saw the\nfollowing behavior:\n\n### NFS v4.0 (stale file case)\n\n![NFS v4.0 flow](https://about.gitlab.com/images/blogimages/nfs-debug/nfs-4.0-flow.svg)\n\nIn this diagram, we can see in step 1 Alice opens `test1.txt` and gets\nback an NFS file handle along with a `stateid` of 0x3000. When Bob\nattempts to rename the file, the NFS server tells to Bob to retry via\nthe `NFS4ERR_DELAY` message while it recalls the delegation from Alice\nvia the `CB_RECALL` message (step 3). Alice then returns her delegation\nvia `DELEGRETURN` (step 4), and then Bob attempts to send another\n`RENAME` message (step 5). The `RENAME` completes in both cases, but\nAlice continues to read using the same file handle.\n\n### NFS v4.1 (working case)\n\n![NFS v4.1 flow](https://about.gitlab.com/images/blogimages/nfs-debug/nfs-4.1-flow.svg)\n\nThe main difference happens at the bottom at step 6. Notice in NFS v4.0\n(the stale file case), Alice attempts to reuse the same `stateid`. In\nNFS v4.1 (working case), Alice performs an additional `LOOKUP` and\n`OPEN`, which causes the server to return a different `stateid`. In v4.0,\nthese extra messages are never sent. This explains why Alice continues\nto see stale content because she uses the old file handle.\n\nWhat makes Alice decide to do the extra `LOOKUP`? The delegation recall\nseemed to work fine, but perhaps there was still an issue, such as a\nmissing invalidation step. To rule that out, we disabled NFS delegations\nby issuing this command on the NFS server itself:\n\n```sh\necho 0 > /proc/sys/fs/leases-enable\n```\n\nWe repeated the experiment, but the problem persisted. All this\nconvinced us this wasn't a NFS server issue or a problem with NFS\ndelegations; it was a problem that led us to look into the NFS client\nwithin the kernel.\n\n## Digging deeper: the Linux NFS client\n\nThe first question we had to answer for the NFS maintainers:\n\n### Was this problem still in the latest upstream kernel?\n\nThe issue occurred with both CentOS 7.2 and Ubuntu 16.04 kernels, which\nused versions 3.10.0-862.11.6 and 4.4.0-130, respectively. However, both\nthose kernels lagged the most recent kernel, which was 4.19-rc2 at the\ntime.\n\nWe deployed a new Ubuntu 16.04 virtual machine on Google Cloud Platform\n(GCP), cloned the latest Linux kernel, and set up a kernel development\nenvironment. After generating a `.config` file via `make menuconfig`, we\nchecked two items:\n\n1. The NFS driver was compiled as a module (`CONFIG_NFSD=m`).\n2. The [required GCP kernel settings](https://cloud.google.com/compute/docs/images/building-custom-os)\nwere set properly.\n\nJust as a geneticist would use fruit flies to study evolution in\nreal time, the first item allowed us to make quick changes in the NFS\nclient without having to reboot the kernel. The second item was required\nto ensure that the kernel would actually boot after it was\ninstalled. Fortunately, the default kernel settings had all the settings\nright out of the box.\n\nWith our custom kernel, we verified that the stale file problem still\nexisted in the latest version. That begged a number of questions:\n\n1. Where exactly was this problem happening?\n2. Why was this problem happening with NFS v4.0 but not in v4.1?\n\nTo answer these questions, we began to investigate the NFS [source\ncode](/solutions/source-code-management/). Since we didn't have a kernel debugger available, we sprinkled the\nsource code with two main types of calls:\n\n1. `pr_info()` ([what used to be `printk`](https://lwn.net/Articles/487437/)).\n2. `dump_stack()`: This would show the stack trace of the current function call.\n\nFor example, one of the first things we did was hook into the\n`nfs4_file_open()` function in `fs/nfs/nfs4file.c`:\n\n```c\nstatic int\nnfs4_file_open(struct inode *inode, struct file *filp)\n{\n...\n        pr_info(\"nfs4_file_open start\\n\");\n        dump_stack();\n\n```\n\nAdmittedly, we could have [activated the `dprintk` messages with the\nLinux dynamic\ndebug](https://www.kernel.org/doc/html/v4.15/admin-guide/dynamic-debug-howto.html)\nor used\n[`rpcdebug`](https://www.thegeekdiary.com/how-to-enable-nfs-debug-logging-using-rpcdebug/),\nbut it was nice to be able to add our own messages to verify changes\nwere being made.\n\nEvery time we made changes, we recompiled the module and reinstalled it\ninto the kernel via the commands:\n\n```sh\nmake modules\nsudo umount /mnt/nfs-test\nsudo rmmod nfsv4\nsudo rmmod nfs\nsudo insmod fs/nfs/nfs.ko\nsudo mount -a\n```\n\nWith our NFS module installed, repeating the experiments would print\nmessages that would help us understand the NFS code a bit more. For\nexample, you can see exactly what happens when an application calls `open()`:\n\n```text\nSep 24 20:20:38 test-kernel kernel: [ 1145.233460] Call Trace:\nSep 24 20:20:38 test-kernel kernel: [ 1145.233462]  dump_stack+0x8e/0xd5\nSep 24 20:20:38 test-kernel kernel: [ 1145.233480] nfs4_file_open+0x56/0x2a0 [nfsv4]\nSep 24 20:20:38 test-kernel kernel: [ 1145.233488]  ? nfs42_clone_file_range+0x1c0/0x1c0 [nfsv4]\nSep 24 20:20:38 test-kernel kernel: [ 1145.233490] do_dentry_open+0x1f6/0x360\nSep 24 20:20:38 test-kernel kernel: [ 1145.233492]  vfs_open+0x2f/0x40\nSep 24 20:20:38 test-kernel kernel: [ 1145.233493]  path_openat+0x2e8/0x1690\nSep 24 20:20:38 test-kernel kernel: [ 1145.233496]  ? mem_cgroup_try_charge+0x8b/0x190\nSep 24 20:20:38 test-kernel kernel: [ 1145.233497]  do_filp_open+0x9b/0x110\nSep 24 20:20:38 test-kernel kernel: [ 1145.233499]  ? __check_object_size+0xb8/0x1b0\nSep 24 20:20:38 test-kernel kernel: [ 1145.233501]  ? __alloc_fd+0x46/0x170\nSep 24 20:20:38 test-kernel kernel: [ 1145.233503]  do_sys_open+0x1ba/0x250\nSep 24 20:20:38 test-kernel kernel: [ 1145.233505]  ? do_sys_open+0x1ba/0x250\nSep 24 20:20:38 test-kernel kernel: [ 1145.233507] __x64_sys_openat+0x20/0x30\nSep 24 20:20:38 test-kernel kernel: [ 1145.233508]  do_syscall_64+0x65/0x130\n```\n\nWhat are the `do_dentry_open` and `vfs_open` calls above? Linux has a\n[virtual filesystem\n(VFS)](https://www.kernel.org/doc/Documentation/filesystems/vfs.txt), an\nabstraction layer which provides a common interface for all\nfilesystems. The VFS documentation explains:\n\n> The VFS implements the open(2), stat(2), chmod(2), and similar system\n> calls. The pathname argument that is passed to them is used by the VFS\n> to search through the directory entry cache (also known as the dentry\n> cache or dcache). This provides a very fast look-up mechanism to\n> translate a pathname (filename) into a specific dentry. Dentries live\n> in RAM and are never saved to disc: they exist only for performance.\n\n### This gave us a clue: what if this was a problem with the dentry cache?\n\nWe noticed a lot of dentry cache validation was done in\n`fs/nfs/dir.c`. In particular, `nfs4_lookup_revalidate()` sounded\npromising. As an experiment, we hacked that function to bail\nout early:\n\n\n```diff\ndiff --git a/fs/nfs/dir.c b/fs/nfs/dir.c\nindex 8bfaa658b2c1..ad479bfeb669 100644\n--- a/fs/nfs/dir.c\n+++ b/fs/nfs/dir.c\n@@ -1159,6 +1159,7 @@ static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)\n        trace_nfs_lookup_revalidate_enter(dir, dentry, flags);\n        error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);\n        trace_nfs_lookup_revalidate_exit(dir, dentry, flags, error);\n+       goto out_bad;\n        if (error == -ESTALE || error == -ENOENT)\n                goto out_bad;\n        if (error)\n\n```\n\nThat made the stale file problem in our experiment go away! Now we were onto something.\n\nTo answer, \"Why does this problem not happen in NFS v4.1?\", we added\n`pr_info()` calls to every `if` block in that function. After running our\nexperiments with NFS v4.0 and v4.1, we found this special condition being run\nin the v4.1 case:\n\n```c\n\n        if (NFS_SB(dentry->d_sb)->caps & NFS_CAP_ATOMIC_OPEN_V1) {\n          goto no_open;\n        }\n\n```\n\nWhat is `NFS_CAP_ATOMIC_OPEN_V1`? We saw [this kernel\npatch](https://patchwork.kernel.org/patch/2300511/) mentioned this was\nan NFS v4.1-specific feature, and the code in `fs/nfs/nfs4proc.c`\nconfirmed that this flag was a capability present in v4.1 but not in v4.0:\n\n```c\nstatic const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {\n        .minor_version = 1,\n        .init_caps = NFS_CAP_READDIRPLUS\n                | NFS_CAP_ATOMIC_OPEN\n                | NFS_CAP_POSIX_LOCK\n                | NFS_CAP_STATEID_NFSV41\n                | NFS_CAP_ATOMIC_OPEN_V1\n\n```\n\nThat explained the difference in behavior: in the v4.1 case, the `goto\nno_open` would cause more validation to happen in\n`nfs_lookup_revalidate()`, but in v4.0, the `nfs4_lookup_revalidate()`\nwould return earlier. Now, how do we actually solve the problem?\n\n## The solution\n\nI reported the [findings to the NFS mailing\nlist](https://marc.info/?l=linux-nfs&m=153782129412452&w=2) and proposed\n[a naive patch](https://marc.info/?l=linux-nfs&m=153807208928650&w=2). A\nweek after the report, Trond Myklebust sent a [patch series to the list\nfixing this bug and found another related issue for NFS\nv4.1](https://marc.info/?l=linux-nfs&m=153816500525563&w=2).\n\nIt turns out the fix for the NFS v4.0 bug was deeper in the code base\nthan we had looked. Trond summarized it well in the\n[patch](https://marc.info/?l=linux-nfs&m=153816500525564&w=2):\n\n> We need to ensure that inode and dentry revalidation occurs correctly\n> on reopen of a file that is already open. Currently, we can end up not\n> revalidating either in the case of NFSv4.0, due to the 'cached open'\n> path.  Let's fix that by ensuring that we only do cached open for the\n> special cases of open recovery and delegation return.\n\nWe confirmed that this fix made the stale file problem go away and filed\nbug reports with\n[Ubuntu](https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1802585)\nand [RedHat](https://bugzilla.redhat.com/show_bug.cgi?id=1648482).\n\nKnowing full well that kernel changes may take a while to make it to\nstable releases, we also added a [workaround in\nGitaly](https://gitlab.com/gitlab-org/gitaly/merge_requests/924) to deal\nwith this issue. We did experiments to test that calling `stat()` on the\n`packed-refs` file appears to cause the kernel to revalidate the dentry\ncache for the renamed file. For simplicity, this is implemented in\nGitaly regardless of whether the filesystem is NFS; we only do this once\nbefore Gitaly \"opens\" a repository, and there are already other `stat()`\ncalls that check for other files.\n\n## What we learned\n\nA bug can be anywhere in your software stack, and sometimes you have to\nlook beyond your application to find it. Having helpful partners in the\nopen source world makes that job much easier.\n\nWe are extremely grateful to Trond Myklebust for fixing the problem, and\nBruce Fields for responding to questions and helping us understand\nNFS. Their responsiveness and professionalism truly reflects the best of\nthe open source community.\n\nPhoto by [dynamosquito](https://www.flickr.com/photos/dynamosquito) on [Flickr](https://www.flickr.com/photos/dynamosquito/4265771518)\n",[23,24,25,26],"community","git","inside GitLab","open source","yml",{},true,"/en-us/blog/how-we-spent-two-weeks-hunting-an-nfs-bug",{"title":15,"description":16,"ogTitle":15,"ogDescription":16,"noIndex":12,"ogImage":19,"ogUrl":32,"ogSiteName":33,"ogType":34,"canonicalUrls":32},"https://about.gitlab.com/blog/how-we-spent-two-weeks-hunting-an-nfs-bug","https://about.gitlab.com","article","en-us/blog/how-we-spent-two-weeks-hunting-an-nfs-bug",[23,24,37,38],"inside-gitlab","open-source","dBxkLN0Vzr8zSsU_M5C2T1U0k_eqSIUgyIfaMj-fQuI",{"data":41},{"logo":42,"freeTrial":47,"sales":52,"login":57,"items":62,"search":369,"minimal":400,"duo":419,"pricingDeployment":429},{"config":43},{"href":44,"dataGaName":45,"dataGaLocation":46},"/","gitlab logo","header",{"text":48,"config":49},"Get free trial",{"href":50,"dataGaName":51,"dataGaLocation":46},"https://gitlab.com/-/trial_registrations/new?glm_source=about.gitlab.com&glm_content=default-saas-trial/","free trial",{"text":53,"config":54},"Talk to sales",{"href":55,"dataGaName":56,"dataGaLocation":46},"/sales/","sales",{"text":58,"config":59},"Sign in",{"href":60,"dataGaName":61,"dataGaLocation":46},"https://gitlab.com/users/sign_in/","sign in",[63,90,185,190,290,350],{"text":64,"config":65,"cards":67},"Platform",{"dataNavLevelOne":66},"platform",[68,74,82],{"title":64,"description":69,"link":70},"The intelligent orchestration platform for DevSecOps",{"text":71,"config":72},"Explore our Platform",{"href":73,"dataGaName":66,"dataGaLocation":46},"/platform/",{"title":75,"description":76,"link":77},"GitLab Duo Agent Platform","Agentic AI for the entire software lifecycle",{"text":78,"config":79},"Meet GitLab Duo",{"href":80,"dataGaName":81,"dataGaLocation":46},"/gitlab-duo-agent-platform/","gitlab duo agent platform",{"title":83,"description":84,"link":85},"Why GitLab","See the top reasons enterprises choose GitLab",{"text":86,"config":87},"Learn more",{"href":88,"dataGaName":89,"dataGaLocation":46},"/why-gitlab/","why gitlab",{"text":91,"left":29,"config":92,"link":94,"lists":98,"footer":167},"Product",{"dataNavLevelOne":93},"solutions",{"text":95,"config":96},"View all Solutions",{"href":97,"dataGaName":93,"dataGaLocation":46},"/solutions/",[99,123,146],{"title":100,"description":101,"link":102,"items":107},"Automation","CI/CD and automation to accelerate deployment",{"config":103},{"icon":104,"href":105,"dataGaName":106,"dataGaLocation":46},"AutomatedCodeAlt","/solutions/delivery-automation/","automated software delivery",[108,112,115,119],{"text":109,"config":110},"CI/CD",{"href":111,"dataGaLocation":46,"dataGaName":109},"/solutions/continuous-integration/",{"text":75,"config":113},{"href":80,"dataGaLocation":46,"dataGaName":114},"gitlab duo agent platform - product menu",{"text":116,"config":117},"Source Code Management",{"href":118,"dataGaLocation":46,"dataGaName":116},"/solutions/source-code-management/",{"text":120,"config":121},"Automated Software Delivery",{"href":105,"dataGaLocation":46,"dataGaName":122},"Automated software delivery",{"title":124,"description":125,"link":126,"items":131},"Security","Deliver code faster without compromising security",{"config":127},{"href":128,"dataGaName":129,"dataGaLocation":46,"icon":130},"/solutions/application-security-testing/","security and compliance","ShieldCheckLight",[132,136,141],{"text":133,"config":134},"Application Security Testing",{"href":128,"dataGaName":135,"dataGaLocation":46},"Application security testing",{"text":137,"config":138},"Software Supply Chain Security",{"href":139,"dataGaLocation":46,"dataGaName":140},"/solutions/supply-chain/","Software supply chain security",{"text":142,"config":143},"Software Compliance",{"href":144,"dataGaName":145,"dataGaLocation":46},"/solutions/software-compliance/","software compliance",{"title":147,"link":148,"items":153},"Measurement",{"config":149},{"icon":150,"href":151,"dataGaName":152,"dataGaLocation":46},"DigitalTransformation","/solutions/visibility-measurement/","visibility and measurement",[154,158,162],{"text":155,"config":156},"Visibility & Measurement",{"href":151,"dataGaLocation":46,"dataGaName":157},"Visibility and Measurement",{"text":159,"config":160},"Value Stream Management",{"href":161,"dataGaLocation":46,"dataGaName":159},"/solutions/value-stream-management/",{"text":163,"config":164},"Analytics & Insights",{"href":165,"dataGaLocation":46,"dataGaName":166},"/solutions/analytics-and-insights/","Analytics and insights",{"title":168,"items":169},"GitLab for",[170,175,180],{"text":171,"config":172},"Enterprise",{"href":173,"dataGaLocation":46,"dataGaName":174},"/enterprise/","enterprise",{"text":176,"config":177},"Small Business",{"href":178,"dataGaLocation":46,"dataGaName":179},"/small-business/","small business",{"text":181,"config":182},"Public Sector",{"href":183,"dataGaLocation":46,"dataGaName":184},"/solutions/public-sector/","public sector",{"text":186,"config":187},"Pricing",{"href":188,"dataGaName":189,"dataGaLocation":46,"dataNavLevelOne":189},"/pricing/","pricing",{"text":191,"config":192,"link":194,"lists":198,"feature":277},"Resources",{"dataNavLevelOne":193},"resources",{"text":195,"config":196},"View all resources",{"href":197,"dataGaName":193,"dataGaLocation":46},"/resources/",[199,232,250],{"title":200,"items":201},"Getting started",[202,207,212,217,222,227],{"text":203,"config":204},"Install",{"href":205,"dataGaName":206,"dataGaLocation":46},"/install/","install",{"text":208,"config":209},"Quick start guides",{"href":210,"dataGaName":211,"dataGaLocation":46},"/get-started/","quick setup checklists",{"text":213,"config":214},"Learn",{"href":215,"dataGaLocation":46,"dataGaName":216},"https://university.gitlab.com/","learn",{"text":218,"config":219},"Product documentation",{"href":220,"dataGaName":221,"dataGaLocation":46},"https://docs.gitlab.com/","product documentation",{"text":223,"config":224},"Best practice videos",{"href":225,"dataGaName":226,"dataGaLocation":46},"/getting-started-videos/","best practice videos",{"text":228,"config":229},"Integrations",{"href":230,"dataGaName":231,"dataGaLocation":46},"/integrations/","integrations",{"title":233,"items":234},"Discover",[235,240,245],{"text":236,"config":237},"Customer success stories",{"href":238,"dataGaName":239,"dataGaLocation":46},"/customers/","customer success stories",{"text":241,"config":242},"Blog",{"href":243,"dataGaName":244,"dataGaLocation":46},"/blog/","blog",{"text":246,"config":247},"Remote",{"href":248,"dataGaName":249,"dataGaLocation":46},"https://handbook.gitlab.com/handbook/company/culture/all-remote/","remote",{"title":251,"items":252},"Connect",[253,258,262,267,272],{"text":254,"config":255},"GitLab Services",{"href":256,"dataGaName":257,"dataGaLocation":46},"/services/","services",{"text":259,"config":260},"Community",{"href":261,"dataGaName":23,"dataGaLocation":46},"/community/",{"text":263,"config":264},"Forum",{"href":265,"dataGaName":266,"dataGaLocation":46},"https://forum.gitlab.com/","forum",{"text":268,"config":269},"Events",{"href":270,"dataGaName":271,"dataGaLocation":46},"/events/","events",{"text":273,"config":274},"Partners",{"href":275,"dataGaName":276,"dataGaLocation":46},"/partners/","partners",{"backgroundColor":278,"textColor":279,"text":280,"image":281,"link":285},"#2f2a6b","#fff","Insights for the future of software development",{"altText":282,"config":283},"the source promo card",{"src":284},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1758208064/dzl0dbift9xdizyelkk4.svg",{"text":286,"config":287},"Read the latest",{"href":288,"dataGaName":289,"dataGaLocation":46},"/the-source/","the source",{"text":291,"config":292,"lists":294},"Company",{"dataNavLevelOne":293},"company",[295],{"items":296},[297,302,308,310,315,320,325,330,335,340,345],{"text":298,"config":299},"About",{"href":300,"dataGaName":301,"dataGaLocation":46},"/company/","about",{"text":303,"config":304,"footerGa":307},"Jobs",{"href":305,"dataGaName":306,"dataGaLocation":46},"/jobs/","jobs",{"dataGaName":306},{"text":268,"config":309},{"href":270,"dataGaName":271,"dataGaLocation":46},{"text":311,"config":312},"Leadership",{"href":313,"dataGaName":314,"dataGaLocation":46},"/company/team/e-group/","leadership",{"text":316,"config":317},"Team",{"href":318,"dataGaName":319,"dataGaLocation":46},"/company/team/","team",{"text":321,"config":322},"Handbook",{"href":323,"dataGaName":324,"dataGaLocation":46},"https://handbook.gitlab.com/","handbook",{"text":326,"config":327},"Investor relations",{"href":328,"dataGaName":329,"dataGaLocation":46},"https://ir.gitlab.com/","investor relations",{"text":331,"config":332},"Trust Center",{"href":333,"dataGaName":334,"dataGaLocation":46},"/security/","trust center",{"text":336,"config":337},"AI Transparency Center",{"href":338,"dataGaName":339,"dataGaLocation":46},"/ai-transparency-center/","ai transparency center",{"text":341,"config":342},"Newsletter",{"href":343,"dataGaName":344,"dataGaLocation":46},"/company/contact/#contact-forms","newsletter",{"text":346,"config":347},"Press",{"href":348,"dataGaName":349,"dataGaLocation":46},"/press/","press",{"text":351,"config":352,"lists":353},"Contact us",{"dataNavLevelOne":293},[354],{"items":355},[356,359,364],{"text":53,"config":357},{"href":55,"dataGaName":358,"dataGaLocation":46},"talk to sales",{"text":360,"config":361},"Support portal",{"href":362,"dataGaName":363,"dataGaLocation":46},"https://support.gitlab.com","support portal",{"text":365,"config":366},"Customer portal",{"href":367,"dataGaName":368,"dataGaLocation":46},"https://customers.gitlab.com/customers/sign_in/","customer portal",{"close":370,"login":371,"suggestions":378},"Close",{"text":372,"link":373},"To search repositories and projects, login to",{"text":374,"config":375},"gitlab.com",{"href":60,"dataGaName":376,"dataGaLocation":377},"search login","search",{"text":379,"default":380},"Suggestions",[381,383,387,389,393,397],{"text":75,"config":382},{"href":80,"dataGaName":75,"dataGaLocation":377},{"text":384,"config":385},"Code Suggestions (AI)",{"href":386,"dataGaName":384,"dataGaLocation":377},"/solutions/code-suggestions/",{"text":109,"config":388},{"href":111,"dataGaName":109,"dataGaLocation":377},{"text":390,"config":391},"GitLab on AWS",{"href":392,"dataGaName":390,"dataGaLocation":377},"/partners/technology-partners/aws/",{"text":394,"config":395},"GitLab on Google Cloud",{"href":396,"dataGaName":394,"dataGaLocation":377},"/partners/technology-partners/google-cloud-platform/",{"text":398,"config":399},"Why GitLab?",{"href":88,"dataGaName":398,"dataGaLocation":377},{"freeTrial":401,"mobileIcon":406,"desktopIcon":411,"secondaryButton":414},{"text":402,"config":403},"Start free trial",{"href":404,"dataGaName":51,"dataGaLocation":405},"https://gitlab.com/-/trials/new/","nav",{"altText":407,"config":408},"Gitlab Icon",{"src":409,"dataGaName":410,"dataGaLocation":405},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1758203874/jypbw1jx72aexsoohd7x.svg","gitlab icon",{"altText":407,"config":412},{"src":413,"dataGaName":410,"dataGaLocation":405},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1758203875/gs4c8p8opsgvflgkswz9.svg",{"text":415,"config":416},"Get Started",{"href":417,"dataGaName":418,"dataGaLocation":405},"https://gitlab.com/-/trial_registrations/new?glm_source=about.gitlab.com/get-started/","get started",{"freeTrial":420,"mobileIcon":425,"desktopIcon":427},{"text":421,"config":422},"Learn more about GitLab Duo",{"href":423,"dataGaName":424,"dataGaLocation":405},"/gitlab-duo/","gitlab duo",{"altText":407,"config":426},{"src":409,"dataGaName":410,"dataGaLocation":405},{"altText":407,"config":428},{"src":413,"dataGaName":410,"dataGaLocation":405},{"freeTrial":430,"mobileIcon":435,"desktopIcon":437},{"text":431,"config":432},"Back to pricing",{"href":188,"dataGaName":433,"dataGaLocation":405,"icon":434},"back to pricing","GoBack",{"altText":407,"config":436},{"src":409,"dataGaName":410,"dataGaLocation":405},{"altText":407,"config":438},{"src":413,"dataGaName":410,"dataGaLocation":405},{"title":440,"button":441,"config":446},"See how agentic AI transforms software delivery",{"text":442,"config":443},"Watch GitLab Transcend now",{"href":444,"dataGaName":445,"dataGaLocation":46},"/events/transcend/virtual/","transcend event",{"layout":447,"icon":448},"release","AiStar",{"data":450},{"text":451,"source":452,"edit":458,"contribute":463,"config":468,"items":473,"minimal":680},"Git is a trademark of Software Freedom Conservancy and our use of 'GitLab' is under license",{"text":453,"config":454},"View page source",{"href":455,"dataGaName":456,"dataGaLocation":457},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/","page source","footer",{"text":459,"config":460},"Edit this page",{"href":461,"dataGaName":462,"dataGaLocation":457},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/-/blob/main/content/","web ide",{"text":464,"config":465},"Please contribute",{"href":466,"dataGaName":467,"dataGaLocation":457},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/-/blob/main/CONTRIBUTING.md/","please contribute",{"twitter":469,"facebook":470,"youtube":471,"linkedin":472},"https://twitter.com/gitlab","https://www.facebook.com/gitlab","https://www.youtube.com/channel/UCnMGQ8QHMAnVIsI3xJrihhg","https://www.linkedin.com/company/gitlab-com",[474,521,575,619,646],{"title":186,"links":475,"subMenu":490},[476,480,485],{"text":477,"config":478},"View plans",{"href":188,"dataGaName":479,"dataGaLocation":457},"view plans",{"text":481,"config":482},"Why Premium?",{"href":483,"dataGaName":484,"dataGaLocation":457},"/pricing/premium/","why premium",{"text":486,"config":487},"Why Ultimate?",{"href":488,"dataGaName":489,"dataGaLocation":457},"/pricing/ultimate/","why ultimate",[491],{"title":492,"links":493},"Contact Us",[494,497,499,501,506,511,516],{"text":495,"config":496},"Contact sales",{"href":55,"dataGaName":56,"dataGaLocation":457},{"text":360,"config":498},{"href":362,"dataGaName":363,"dataGaLocation":457},{"text":365,"config":500},{"href":367,"dataGaName":368,"dataGaLocation":457},{"text":502,"config":503},"Status",{"href":504,"dataGaName":505,"dataGaLocation":457},"https://status.gitlab.com/","status",{"text":507,"config":508},"Terms of use",{"href":509,"dataGaName":510,"dataGaLocation":457},"/terms/","terms of use",{"text":512,"config":513},"Privacy statement",{"href":514,"dataGaName":515,"dataGaLocation":457},"/privacy/","privacy statement",{"text":517,"config":518},"Cookie preferences",{"dataGaName":519,"dataGaLocation":457,"id":520,"isOneTrustButton":29},"cookie preferences","ot-sdk-btn",{"title":91,"links":522,"subMenu":531},[523,527],{"text":524,"config":525},"DevSecOps platform",{"href":73,"dataGaName":526,"dataGaLocation":457},"devsecops platform",{"text":528,"config":529},"AI-Assisted Development",{"href":423,"dataGaName":530,"dataGaLocation":457},"ai-assisted development",[532],{"title":533,"links":534},"Topics",[535,540,545,550,555,560,565,570],{"text":536,"config":537},"CICD",{"href":538,"dataGaName":539,"dataGaLocation":457},"/topics/ci-cd/","cicd",{"text":541,"config":542},"GitOps",{"href":543,"dataGaName":544,"dataGaLocation":457},"/topics/gitops/","gitops",{"text":546,"config":547},"DevOps",{"href":548,"dataGaName":549,"dataGaLocation":457},"/topics/devops/","devops",{"text":551,"config":552},"Version Control",{"href":553,"dataGaName":554,"dataGaLocation":457},"/topics/version-control/","version control",{"text":556,"config":557},"DevSecOps",{"href":558,"dataGaName":559,"dataGaLocation":457},"/topics/devsecops/","devsecops",{"text":561,"config":562},"Cloud Native",{"href":563,"dataGaName":564,"dataGaLocation":457},"/topics/cloud-native/","cloud native",{"text":566,"config":567},"AI for Coding",{"href":568,"dataGaName":569,"dataGaLocation":457},"/topics/devops/ai-for-coding/","ai for coding",{"text":571,"config":572},"Agentic AI",{"href":573,"dataGaName":574,"dataGaLocation":457},"/topics/agentic-ai/","agentic ai",{"title":576,"links":577},"Solutions",[578,580,582,587,591,594,598,601,603,606,609,614],{"text":133,"config":579},{"href":128,"dataGaName":133,"dataGaLocation":457},{"text":122,"config":581},{"href":105,"dataGaName":106,"dataGaLocation":457},{"text":583,"config":584},"Agile development",{"href":585,"dataGaName":586,"dataGaLocation":457},"/solutions/agile-delivery/","agile delivery",{"text":588,"config":589},"SCM",{"href":118,"dataGaName":590,"dataGaLocation":457},"source code management",{"text":536,"config":592},{"href":111,"dataGaName":593,"dataGaLocation":457},"continuous integration & delivery",{"text":595,"config":596},"Value stream management",{"href":161,"dataGaName":597,"dataGaLocation":457},"value stream management",{"text":541,"config":599},{"href":600,"dataGaName":544,"dataGaLocation":457},"/solutions/gitops/",{"text":171,"config":602},{"href":173,"dataGaName":174,"dataGaLocation":457},{"text":604,"config":605},"Small business",{"href":178,"dataGaName":179,"dataGaLocation":457},{"text":607,"config":608},"Public sector",{"href":183,"dataGaName":184,"dataGaLocation":457},{"text":610,"config":611},"Education",{"href":612,"dataGaName":613,"dataGaLocation":457},"/solutions/education/","education",{"text":615,"config":616},"Financial services",{"href":617,"dataGaName":618,"dataGaLocation":457},"/solutions/finance/","financial services",{"title":191,"links":620},[621,623,625,627,630,632,634,636,638,640,642,644],{"text":203,"config":622},{"href":205,"dataGaName":206,"dataGaLocation":457},{"text":208,"config":624},{"href":210,"dataGaName":211,"dataGaLocation":457},{"text":213,"config":626},{"href":215,"dataGaName":216,"dataGaLocation":457},{"text":218,"config":628},{"href":220,"dataGaName":629,"dataGaLocation":457},"docs",{"text":241,"config":631},{"href":243,"dataGaName":244,"dataGaLocation":457},{"text":236,"config":633},{"href":238,"dataGaName":239,"dataGaLocation":457},{"text":246,"config":635},{"href":248,"dataGaName":249,"dataGaLocation":457},{"text":254,"config":637},{"href":256,"dataGaName":257,"dataGaLocation":457},{"text":259,"config":639},{"href":261,"dataGaName":23,"dataGaLocation":457},{"text":263,"config":641},{"href":265,"dataGaName":266,"dataGaLocation":457},{"text":268,"config":643},{"href":270,"dataGaName":271,"dataGaLocation":457},{"text":273,"config":645},{"href":275,"dataGaName":276,"dataGaLocation":457},{"title":291,"links":647},[648,650,652,654,656,658,660,664,669,671,673,675],{"text":298,"config":649},{"href":300,"dataGaName":293,"dataGaLocation":457},{"text":303,"config":651},{"href":305,"dataGaName":306,"dataGaLocation":457},{"text":311,"config":653},{"href":313,"dataGaName":314,"dataGaLocation":457},{"text":316,"config":655},{"href":318,"dataGaName":319,"dataGaLocation":457},{"text":321,"config":657},{"href":323,"dataGaName":324,"dataGaLocation":457},{"text":326,"config":659},{"href":328,"dataGaName":329,"dataGaLocation":457},{"text":661,"config":662},"Sustainability",{"href":663,"dataGaName":661,"dataGaLocation":457},"/sustainability/",{"text":665,"config":666},"Diversity, inclusion and belonging (DIB)",{"href":667,"dataGaName":668,"dataGaLocation":457},"/diversity-inclusion-belonging/","Diversity, inclusion and belonging",{"text":331,"config":670},{"href":333,"dataGaName":334,"dataGaLocation":457},{"text":341,"config":672},{"href":343,"dataGaName":344,"dataGaLocation":457},{"text":346,"config":674},{"href":348,"dataGaName":349,"dataGaLocation":457},{"text":676,"config":677},"Modern Slavery Transparency Statement",{"href":678,"dataGaName":679,"dataGaLocation":457},"https://handbook.gitlab.com/handbook/legal/modern-slavery-act-transparency-statement/","modern slavery transparency statement",{"items":681},[682,685,688],{"text":683,"config":684},"Terms",{"href":509,"dataGaName":510,"dataGaLocation":457},{"text":686,"config":687},"Cookies",{"dataGaName":519,"dataGaLocation":457,"id":520,"isOneTrustButton":29},{"text":689,"config":690},"Privacy",{"href":514,"dataGaName":515,"dataGaLocation":457},[692],{"id":693,"title":18,"body":8,"config":694,"content":696,"description":8,"extension":27,"meta":700,"navigation":29,"path":701,"seo":702,"stem":703,"__hash__":704},"blogAuthors/en-us/blog/authors/stan-hu.yml",{"template":695},"BlogAuthor",{"name":18,"config":697},{"headshot":698,"ctfId":699},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1749659504/Blog/Author%20Headshots/stanhu-headshot.jpg","stanhu",{},"/en-us/blog/authors/stan-hu",{},"en-us/blog/authors/stan-hu","KmQVCb_7YcWghHApaS2EI3J2bQ0dRustgOz4wYyOnVk",[706,721,733],{"content":707,"config":719},{"title":708,"description":709,"authors":710,"heroImage":712,"date":713,"body":714,"category":9,"tags":715},"How to use GitLab Container Virtual Registry with Docker Hardened Images","Learn how to simplify container image management with this step-by-step guide.",[711],"Tim Rizzi","https://res.cloudinary.com/about-gitlab-com/image/upload/v1772111172/mwhgbjawn62kymfwrhle.png","2026-03-12","If you're a platform engineer, you've probably had this conversation:\n  \n*\"Security says we need to use hardened base images.\"*\n\n*\"Great, where do I configure credentials for yet another registry?\"*\n\n*\"Also, how do we make sure everyone actually uses them?\"*\n\nOr this one:\n\n*\"Why are our builds so slow?\"*\n\n*\"We're pulling the same 500MB image from Docker Hub in every single job.\"*\n\n*\"Can't we just cache these somewhere?\"*\n\nI've been working on [Container Virtual Registry](https://docs.gitlab.com/user/packages/virtual_registry/container/) at GitLab specifically to solve these problems. It's a pull-through cache that sits in front of your upstream registries — Docker Hub, dhi.io (Docker Hardened Images), MCR, and Quay — and gives your teams a single endpoint to pull from. Images get cached on the first pull. Subsequent pulls come from the cache. Your developers don't need to know or care which upstream a particular image came from.\n\nThis article shows you how to set up Container Virtual Registry, specifically with Docker Hardened Images in mind, since that's a combination that makes a lot of sense for teams concerned about security and not making their developers' lives harder.\n\n## What problem are we actually solving?\n\nThe Platform teams I usually talk to manage container images across three to five registries:\n\n* **Docker Hub** for most base images\n* **dhi.io** for Docker Hardened Images (security-conscious workloads)\n* **MCR** for .NET and Azure tooling\n* **Quay.io** for Red Hat ecosystem stuff\n* **Internal registries** for proprietary images\n\nEach one has its own:\n\n* Authentication mechanism\n* Network latency characteristics\n* Way of organizing image paths\n\nYour CI/CD configs end up littered with registry-specific logic. Credential management becomes a project unto itself. And every pipeline job pulls the same base images over the network, even though they haven't changed in weeks.\n\nContainer Virtual Registry consolidates this. One registry URL. One authentication flow (GitLab's). Cached images are served from GitLab's infrastructure rather than traversing the internet each time.\n\n## How it works\n\nThe model is straightforward:\n\n```text\nYour pipeline pulls:\n  gitlab.com/virtual_registries/container/1000016/python:3.13\n\nVirtual registry checks:\n  1. Do I have this cached? → Return it\n  2. No? → Fetch from upstream, cache it, return it\n\n```\n\nYou configure upstreams in priority order. When a pull request comes in, the virtual registry checks each upstream until it finds the image. The result gets cached for a configurable period (default 24 hours).\n\n```text\n┌─────────────────────────────────────────────────────────┐\n│                    CI/CD Pipeline                       │\n│                          │                              │\n│                          ▼                              │\n│   gitlab.com/virtual_registries/container/\u003Cid>/image   │\n└─────────────────────────────────────────────────────────┘\n                           │\n                           ▼\n┌─────────────────────────────────────────────────────────┐\n│            Container Virtual Registry                   │\n│                                                         │\n│  Upstream 1: Docker Hub ────────────────┐               │\n│  Upstream 2: dhi.io (Hardened) ────────┐│               │\n│  Upstream 3: MCR ─────────────────────┐││               │\n│  Upstream 4: Quay.io ────────────────┐│││               │\n│                                      ││││               │\n│                    ┌─────────────────┴┴┴┴──┐            │\n│                    │        Cache          │            │\n│                    │  (manifests + layers) │            │\n│                    └───────────────────────┘            │\n└─────────────────────────────────────────────────────────┘\n```\n\n## Why this matters for Docker Hardened Images\n\n[Docker Hardened Images](https://docs.docker.com/dhi/) are great because of the minimal attack surface, near-zero CVEs, proper software bills of materials (SBOMs), and SLSA provenance. If you're evaluating base images for security-sensitive workloads, they should be on your list.\n\nBut adopting them creates the same operational friction as any new registry:\n\n* **Credential distribution**: You need to get Docker credentials to every system that pulls images from dhi.io.\n* **CI/CD changes**: Every pipeline needs to be updated to authenticate with dhi.io.\n* **Developer friction**: People need to remember to use the hardened variants.\n* **Visibility gap**: It's difficulat to tell if teams are actually using hardened images vs. regular ones.\n\nVirtual registry addresses each of these:\n\n**Single credential**: Teams authenticate to GitLab. The virtual registry handles upstream authentication. You configure Docker credentials once, at the registry level, and they apply to all pulls.\n\n**No CI/CD changes per-team**: Point pipelines at your virtual registry. Done. The upstream configuration is centralized.\n\n**Gradual adoption**: Since images get cached with their full path, you can see in the cache what's being pulled. If someone's pulling `library/python:3.11` instead of the hardened variant, you'll know.\n\n**Audit trail**: The cache shows you exactly which images are in active use. Useful for compliance, useful for understanding what your fleet actually depends on.\n\n## Setting it up\n\nHere's a real setup using the Python client from this demo project.\n\n### Create the virtual registry\n\n```python\nfrom virtual_registry_client import VirtualRegistryClient\n\nclient = VirtualRegistryClient()\n\nregistry = client.create_virtual_registry(\n    group_id=\"785414\",  # Your top-level group ID\n    name=\"platform-images\",\n    description=\"Cached container images for platform teams\"\n)\n\nprint(f\"Registry ID: {registry['id']}\")\n# You'll need this ID for the pull URL\n```\n\n### Add Docker Hub as an upstream\n\nFor official images like Alpine, Python, etc.:\n\n```python\ndocker_upstream = client.create_upstream(\n    registry_id=registry['id'],\n    url=\"https://registry-1.docker.io\",\n    name=\"Docker Hub\",\n    cache_validity_hours=24\n)\n```\n\n### Add Docker Hardened Images (dhi.io)\n\nDocker Hardened Images are hosted on `dhi.io`, a separate registry that requires authentication:\n\n```python\ndhi_upstream = client.create_upstream(\n    registry_id=registry['id'],\n    url=\"https://dhi.io\",\n    name=\"Docker Hardened Images\",\n    username=\"your-docker-username\",\n    password=\"your-docker-access-token\",\n    cache_validity_hours=24\n)\n```\n\n### Add other upstreams\n\n```python\n# MCR for .NET teams\nclient.create_upstream(\n    registry_id=registry['id'],\n    url=\"https://mcr.microsoft.com\",\n    name=\"Microsoft Container Registry\",\n    cache_validity_hours=48\n)\n\n# Quay for Red Hat stuff\nclient.create_upstream(\n    registry_id=registry['id'],\n    url=\"https://quay.io\",\n    name=\"Quay.io\",\n    cache_validity_hours=24\n)\n```\n\n### Update your CI/CD\n\nHere's a `.gitlab-ci.yml` that pulls through the virtual registry:\n\n```yaml\nvariables:\n  VIRTUAL_REGISTRY_ID: \u003Cyour_virtual_registry_ID>\n\n  \nbuild:\n  image: docker:24\n  services:\n    - docker:24-dind\n  before_script:\n    # Authenticate to GitLab (which handles upstream auth for you)\n    - echo \"${CI_JOB_TOKEN}\" | docker login -u gitlab-ci-token --password-stdin gitlab.com\n  script:\n    # All of these go through your single virtual registry\n    \n    # Official Docker Hub images (use library/ prefix)\n    - docker pull gitlab.com/virtual_registries/container/${VIRTUAL_REGISTRY_ID}/library/alpine:latest\n    \n    # Docker Hardened Images from dhi.io (no prefix needed)\n    - docker pull gitlab.com/virtual_registries/container/${VIRTUAL_REGISTRY_ID}/python:3.13\n    \n    # .NET from MCR\n    - docker pull gitlab.com/virtual_registries/container/${VIRTUAL_REGISTRY_ID}/dotnet/sdk:8.0\n```\n\n### Image path formats\n\nDifferent registries use different path conventions:\n\n| Registry | Pull URL Example |\n|----------|------------------|\n| Docker Hub (official) | `.../library/python:3.11-slim` |\n| Docker Hardened Images (dhi.io) | `.../python:3.13` |\n| MCR | `.../dotnet/sdk:8.0` |\n| Quay.io | `.../prometheus/prometheus:latest` |\n\n### Verify it's working\n\nAfter some pulls, check your cache:\n\n```python\nupstreams = client.list_registry_upstreams(registry['id'])\nfor upstream in upstreams:\n    entries = client.list_cache_entries(upstream['id'])\n    print(f\"{upstream['name']}: {len(entries)} cached entries\")\n\n```\n\n## What the numbers look like\n\nI ran tests pulling images through the virtual registry:\n\n| Metric | Without Cache | With Warm Cache |\n|--------|---------------|-----------------|\n| Pull time (Alpine) | 10.3s | 4.2s |\n| Pull time (Python 3.13 DHI) | 11.6s | ~4s |\n| Network roundtrips to upstream | Every pull | Cache misses only |\n\n\n\n\nThe first pull is the same speed (it has to fetch from upstream). Every pull after that, for the cache validity period, comes straight from GitLab's storage. No network hop to Docker Hub, dhi.io, MCR, or wherever the image lives.\n\nFor a team running hundreds of pipeline jobs per day, that's hours of cumulative build time saved.\n\n## Practical considerations\nHere are some considerations to keep in mind:\n\n### Cache validity\n\n24 hours is the default. For security-sensitive images where you want patches quickly, consider 12 hours or less:\n\n```python\nclient.create_upstream(\n    registry_id=registry['id'],\n    url=\"https://dhi.io\",\n    name=\"Docker Hardened Images\",\n    username=\"your-username\",\n    password=\"your-token\",\n    cache_validity_hours=12\n)\n```\n\nFor stable, infrequently-updated images (like specific version tags), longer validity is fine.\n\n### Upstream priority\n\nUpstreams are checked in order. If you have images with the same name on different registries, the first matching upstream wins.\n\n### Limits\n\n* Maximum of 20 virtual registries per group\n* Maximum of 20 upstreams per virtual registry\n\n## Configuration via UI\n\nYou can also configure virtual registries and upstreams directly from the GitLab UI—no API calls required. Navigate to your group's **Settings > Packages and registries > Virtual Registry** to:\n\n* Create and manage virtual registries\n* Add, edit, and reorder upstream registries\n* View and manage the cache\n* Monitor which images are being pulled\n\n## What's next\n\nWe're actively developing:\n\n* **Allow/deny lists**: Use regex to control which images can be pulled from specific upstreams.\n\nThis is beta software. It works, people are using it in production, but we're still iterating based on feedback.\n\n## Share your feedback\n\nIf you're a platform engineer dealing with container registry sprawl, I'd like to understand your setup:\n\n* How many upstream registries are you managing?\n* What's your biggest pain point with the current state?\n* Would something like this help, and if not, what's missing?\n\nPlease share your experiences in the [Container Virtual Registry feedback issue](https://gitlab.com/gitlab-org/gitlab/-/work_items/589630).\n## Related resources\n- [New GitLab metrics and registry features help reduce CI/CD bottlenecks](https://about.gitlab.com/blog/new-gitlab-metrics-and-registry-features-help-reduce-ci-cd-bottlenecks/#container-virtual-registry)\n- [Container Virtual Registry documentation](https://docs.gitlab.com/user/packages/virtual_registry/container/)\n- [Container Virtual Registry API](https://docs.gitlab.com/api/container_virtual_registries/)",[716,717,718],"tutorial","product","features",{"featured":12,"template":13,"slug":720},"using-gitlab-container-virtual-registry-with-docker-hardened-images",{"content":722,"config":731},{"title":723,"description":724,"authors":725,"heroImage":727,"date":728,"category":9,"tags":729,"body":730},"How IIT Bombay students are coding the future with GitLab","At GitLab, we often talk about how software accelerates innovation. But sometimes, you have to step away from the Zoom calls and stand in a crowded university hall to remember why we do this.",[726],"Nick Veenhof","https://res.cloudinary.com/about-gitlab-com/image/upload/v1750099013/Blog/Hero%20Images/Blog/Hero%20Images/blog-image-template-1800x945%20%2814%29_6VTUA8mUhOZNDaRVNPeKwl_1750099012960.png","2026-01-08",[23,613,26],"The GitLab team recently had the privilege of judging the **iHack Hackathon** at **IIT Bombay's E-Summit**. The energy was electric, the coffee was flowing, and the talent was undeniable. But what struck us most wasn't just the code — it was the sheer determination of students to solve real-world problems, often overcoming significant logistical and financial hurdles to simply be in the room.\n\n\nThrough our [GitLab for Education program](https://about.gitlab.com/solutions/education/), we aim to empower the next generation of developers with tools and opportunity. Here is a look at what the students built, and how they used GitLab to bridge the gap between idea and reality.\n\n## The challenge: Build faster, build securely\n\nThe premise for the GitLab track of the hackathon was simple: Don't just show us a product; show us how you built it. We wanted to see how students utilized GitLab's platform — from Issue Boards to CI/CD pipelines — to accelerate the development lifecycle.\n\nThe results were inspiring.\n\n## The winners\n\n### 1st place: Team Decode — Democratizing Scientific Research\n\n**Project:** FIRE (Fast Integrated Research Environment)\n\nTeam Decode took home the top prize with a solution that warms a developer's heart: a local-first, blazing-fast data processing tool built with [Rust](https://about.gitlab.com/blog/secure-rust-development-with-gitlab/) and Tauri. They identified a massive pain point for data science students: existing tools are fragmented, slow, and expensive.\n\nTheir solution, FIRE, allows researchers to visualize complex formats (like NetCDF) instantly. What impressed the judges most was their \"hacker\" ethos. They didn't just build a tool; they built it to be open and accessible.\n\n**How they used GitLab:** Since the team lived far apart, asynchronous communication was key. They utilized **GitLab Issue Boards** and **Milestones** to track progress and integrated their repo with Telegram to get real-time push notifications. As one team member noted, \"Coordinating all these technologies was really difficult, and what helped us was GitLab... the Issue Board really helped us track who was doing what.\"\n\n![Team Decode](https://res.cloudinary.com/about-gitlab-com/image/upload/v1767380253/epqazj1jc5c7zkgqun9h.jpg)\n\n### 2nd place: Team BichdeHueDost — Reuniting to Solve Payments\n\n**Project:** SemiPay (RFID Cashless Payment for Schools)\n\nThe team name, BichdeHueDost, translates to \"Friends who have been set apart.\" It's a fitting name for a group of friends who went to different colleges but reunited to build this project. They tackled a unique problem: handling cash in schools for young children. Their solution used RFID cards backed by a blockchain ledger to ensure secure, cashless transactions for students.\n\n**How they used GitLab:** They utilized [GitLab CI/CD](https://about.gitlab.com/topics/ci-cd/) to automate the build process for their Flutter application (APK), ensuring that every commit resulted in a testable artifact. This allowed them to iterate quickly despite the \"flaky\" nature of cross-platform mobile development.\n\n![Team BichdeHueDost](https://res.cloudinary.com/about-gitlab-com/image/upload/v1767380253/pkukrjgx2miukb6nrj5g.jpg)\n\n### 3rd place: Team ZenYukti — Agentic Repository Intelligence\n\n**Project:** RepoInsight AI (AI-powered, GitLab-native intelligence platform)\n\nTeam ZenYukti impressed us with a solution that tackles a universal developer pain point: understanding unfamiliar codebases. What stood out to the judges was the tool's practical approach to onboarding and code comprehension: RepoInsight-AI automatically generates documentation, visualizes repository structure, and even helps identify bugs, all while maintaining context about the entire codebase.\n\n**How they used GitLab:** The team built a comprehensive CI/CD pipeline that showcased GitLab's security and DevOps capabilities. They integrated [GitLab's Security Templates](https://gitlab.com/gitlab-org/gitlab/-/tree/master/lib/gitlab/ci/templates/Security) (SAST, Dependency Scanning, and Secret Detection), and utilized [GitLab Container Registry](https://docs.gitlab.com/user/packages/container_registry/) to manage their Docker images for backend and frontend components. They created an AI auto-review bot that runs on merge requests, demonstrating an \"agentic workflow\" where AI assists in the development process itself.\n\n![Team ZenYukti](https://res.cloudinary.com/about-gitlab-com/image/upload/v1767380253/ymlzqoruv5al1secatba.jpg)\n\n## Beyond the code: A lesson in inclusion\n\nWhile the code was impressive, the most powerful moment of the event happened away from the keyboard.\n\nDuring the feedback session, we learned about the journey Team ZenYukti took to get to Mumbai. They traveled over 24 hours, covering nearly 1,800 kilometers. Because flights were too expensive and trains were booked, they traveled in the \"General Coach,\" a non-reserved, severely overcrowded carriage.\n\nAs one student described it:\n\n*\"You cannot even imagine something like this... there are no seats... people sit on the top of the train. This is what we have endured.\"*\n\nThis hit home. [Diversity, Inclusion, and Belonging](https://handbook.gitlab.com/handbook/company/culture/inclusion/) are core values at GitLab. We realized that for these students, the barrier to entry wasn't intellect or skill, it was access.\n\nIn that moment, we decided to break that barrier. We committed to reimbursing the travel expenses for the participants who struggled to get there. It's a small step, but it underlines a massive truth: **talent is distributed equally, but opportunity is not.**\n\n![hackathon class together](https://res.cloudinary.com/about-gitlab-com/image/upload/v1767380252/o5aqmboquz8ehusxvgom.jpg)\n\n### The future is bright (and automated)\n\nWe also saw incredible potential in teams like Prometheus, who attempted to build an autonomous patch remediation tool (DevGuardian), and Team Arrakis, who built a voice-first job portal for blue-collar workers using [GitLab Duo](https://about.gitlab.com/gitlab-duo/) to troubleshoot their pipelines.\n\nTo all the students who participated: You are the future. Through [GitLab for Education](https://about.gitlab.com/solutions/education/), we are committed to providing you with the top-tier tools (like GitLab Ultimate) you need to learn, collaborate, and change the world — whether you are coding from a dorm room, a lab, or a train carriage. **Keep shipping.**\n\n> :bulb: Learn more about the [GitLab for Education program](https://about.gitlab.com/solutions/education/).\n",{"slug":732,"featured":12,"template":13},"how-iit-bombay-students-code-future-with-gitlab",{"content":734,"config":742},{"title":735,"description":736,"authors":737,"heroImage":738,"date":739,"category":9,"tags":740,"body":741},"Artois University elevates research and curriculum with GitLab Ultimate for Education","Artois University's CRIL leveraged the GitLab for Education program to gain free access to Ultimate, transforming advanced research and computer science curricula.",[726],"https://res.cloudinary.com/about-gitlab-com/image/upload/v1750099203/Blog/Hero%20Images/Blog/Hero%20Images/blog-image-template-1800x945%20%2820%29_2bJGC5ZP3WheoqzlLT05C5_1750099203484.png","2025-12-10",[613,23,717],"Leading academic institutions face a critical challenge: how to provide thousands of students and researchers with industry-standard, **full-featured DevSecOps tools** without compromising institutional control. Many start with basic version control, but the modern curriculum demands integrated capabilities for planning, security, and advanced CI/CD.\n\nThe **GitLab for Education program** is designed to solve this by providing access to **GitLab Ultimate** for qualifying institutions, allowing them to scale their operations and elevate their academic offerings. \n\nThis article showcases a powerful success story from the **Centre de Recherche en Informatique de Lens (CRIL)**, a joint laboratory of **Artois University** and CNRS in France. After years of relying solely on GitLab Community Edition (CE), the university's move to GitLab Ultimate through the GitLab for Education program immediately unlocked advanced capabilities, transforming their teaching, research, and contribution workflows virtually overnight. This story demonstrates why GitLab Ultimate is essential for institutions seeking to deliver advanced computer science and research curricula.\n\n## GitLab Ultimate unlocked: Managing scale and driving academic value\n\n**Artois University's** self-managed GitLab instance is a large-scale operation, supporting nearly **3,000 users** across approximately **19,000 projects**, primarily serving computer science students and researchers. While GitLab Community Edition was robust, the upgrade to GitLab Ultimate provided the sophisticated tooling necessary for managing this scale and facilitating advanced university-level work.\n\n***\"We can see the difference,\" says Daniel Le Berre, head of research at CRIL and the instance maintainer. \"It's a completely different product. Each week reveals new features that directly enhance our productivity and teaching.\"***\n\nThe institution joined the GitLab for Education program specifically because it covers both **instructional and non-commercial research use cases** and offers full access to Ultimate's features, removing significant cost barriers.\n\n### Key GitLab Ultimate benefits for students and researchers\n\n* **Advanced project management at scale:** Master's students now benefit from **GitLab Ultimate's project planning features**. This enables them to structure, track, and manage complex, long-term research projects using professional methodologies like portfolio management and advanced issue tracking that seamlessly roll up across their thousands of projects.\n\n* **Enhanced visibility:** Features like improved dashboards and code previews directly in Markdown files dramatically streamline tracking and documentation review, reducing administrative friction for both instructors and students managing large project loads.\n\n## Comprehensive curriculum: From concepts to continuous delivery\n\nGitLab Ultimate is deeply integrated into the computer science curriculum, moving students beyond simple `git` commands to practical **DevSecOps implementation**.\n\n* **Git fundamentals:** Students begin by visualizing concepts using open-source tools to master Git concepts.\n\n* **Full CI/CD implementation:** Students use GitLab CI for rigorous **Test-Driven Development (TDD)** in their software projects. They learn to build, test, and perform quality assurance using unit and integration testing pipelines—core competency made seamless by the integrated platform.\n\n* **DevSecOps for research and documentation:** The university teaches students that DevSecOps principles are vital for all collaborative work. Inspired by earlier work in Delft, students manage and produce critical research documentation (PDFs from Markdown files) using GitLab, incorporating quality checks like linters and spell checks directly in the CI pipeline. This ensures high-quality, reproducible research output.\n\n* **Future-proofing security skills:** The GitLab Ultimate platform immediately positions the institution to incorporate advanced DevSecOps features like SAST and DAST scanning as their research and development code projects grow, ensuring students are prepared for industry security standards.\n\n## Accelerating open source contributions with GitLab Duo\n\nAccess to the full GitLab platform, including our AI capabilities, has empowered students to make impactful contributions to the wider open source community faster than ever before.\n\nTwo Master's students recently completed direct contributions to the GitLab product, adding the **ORCID identifier** into user profiles. Working on GitLab.com, they leveraged **GitLab Duo's AI chat and code suggestions** to navigate the codebase efficiently.\n\n***\"This would not have been possible without GitLab Duo,\" Daniel Le Berre notes. \"The AI features helped students, who might have lacked deep codebase knowledge, deliver meaningful contributions in just two weeks.\"***\n\nThis demonstrates how providing students with cutting-edge tools **accelerates their learning and impact**, allowing them to translate classroom knowledge into real-world contributions immediately.\n\n## Empowering open research and institutional control\n\nThe stability of the self-managed instance at Artois University is key to its success. This model guarantees **institutional control and stability** — a critical factor for long-term research preservation.\n\nThe institution's expertise in this area was recently highlighted in a major 2024 study led by CRIL, titled: \"[Higher Education and Research Forges in France - Definition, uses, limitations encountered and needs analysis](https://hal.science/hal-04208924v4)\" ([Project on GitLab](https://gitlab.in2p3.fr/coso-college-codes-sources-et-logiciels/forges-esr-en)). The research found that the vast majority of public forges in French Higher Education and Research relied on **GitLab**. This finding underscores the consensus among academic leaders that self-hosted solutions are essential for **data control and longevity**, especially when compared to relying on external, commercial forges.\n\n## Unlock GitLab Ultimate for your institution today\n\nThe success story of **Artois University's CRIL** proves the transformative power of the GitLab for Education program. By providing **free access to GitLab Ultimate**, we enable large-scale institutions to:\n\n1.  **Deliver a modern, integrated DevSecOps curriculum.**\n\n2.  **Support advanced, collaborative research projects with Ultimate planning features.**\n\n3.  **Empower students to make AI-assisted open source contributions.**\n\n4.  **Maintain institutional control and data longevity.**\n\nIf your academic institution is ready to equip its students and researchers with the complete DevSecOps platform and its most advanced features, we invite you to join the program.\n\nThe program provides **free access to GitLab Ultimate** for qualifying instructional and non-commercial research use cases.\n\n**Apply now [online](https://about.gitlab.com/solutions/education/join/).**\n",{"slug":743,"featured":29,"template":13},"artois-university-elevates-curriculum-with-gitlab-ultimate-for-education",{"promotions":745},[746,760,771],{"id":747,"categories":748,"header":750,"text":751,"button":752,"image":757},"ai-modernization",[749],"ai-ml","Is AI achieving its promise at scale?","Quiz will take 5 minutes or less",{"text":753,"config":754},"Get your AI maturity score",{"href":755,"dataGaName":756,"dataGaLocation":244},"/assessments/ai-modernization-assessment/","modernization assessment",{"config":758},{"src":759},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1772138786/qix0m7kwnd8x2fh1zq49.png",{"id":761,"categories":762,"header":763,"text":751,"button":764,"image":768},"devops-modernization",[717,559],"Are you just managing tools or shipping innovation?",{"text":765,"config":766},"Get your DevOps maturity score",{"href":767,"dataGaName":756,"dataGaLocation":244},"/assessments/devops-modernization-assessment/",{"config":769},{"src":770},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1772138785/eg818fmakweyuznttgid.png",{"id":772,"categories":773,"header":775,"text":751,"button":776,"image":780},"security-modernization",[774],"security","Are you trading speed for security?",{"text":777,"config":778},"Get your security maturity score",{"href":779,"dataGaName":756,"dataGaLocation":244},"/assessments/security-modernization-assessment/",{"config":781},{"src":782},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1772138786/p4pbqd9nnjejg5ds6mdk.png",{"header":784,"blurb":785,"button":786,"secondaryButton":791},"Start building faster today","See what your team can do with the intelligent orchestration platform for DevSecOps.\n",{"text":787,"config":788},"Get your free trial",{"href":789,"dataGaName":51,"dataGaLocation":790},"https://gitlab.com/-/trial_registrations/new?glm_content=default-saas-trial&glm_source=about.gitlab.com/","feature",{"text":495,"config":792},{"href":55,"dataGaName":56,"dataGaLocation":790},1773350827265]