runbook / 008 · Tailwind Resource Group · Customer advisory · Disclosed November 20, 2026 · Updated November 20, 2026
Upgrade GitLab to 19.6.1, 19.5.8 or 19.4.10 to remove the Ruby flaw CVE-2026-132825
On November 20, 2026, GitLab released 19.6.1, 19.5.8 and 19.4.10. They update the Ruby that GitLab includes, to fix CVE-2026-132825 and the other Ruby flaws of November 18. GitLab says to upgrade all self-managed installations immediately. This page tells you which installations need it and how to confirm the fix.
- Applies to
- Self-managed GitLab CE and EE
- Fix
19.6.1,19.5.8,19.4.10- Severity
- Critical (Ruby), High 7.4 (CVE record)
- GitLab.com
- Already fixed, no action
What changed on November 20, 2026
- OpenShift: how to upgrade with the GitLab Operator. The fixed charts need Operator
3.6.1. We tested the upgrade on OpenShift 4.19. - GitLab Geo: upgrade every site, the primary first, then check each site.
Summary
On November 18, 2026, the Ruby security team disclosed seven flaws in the Ruby regular expression engine and in the case conversion of strings. It released Ruby 4.0.10, 3.4.13 and 3.3.14 to fix them. The Ruby security team says that it knows of proof-of-concept code that uses some of these flaws for remote code execution.
GitLab includes its own Ruby. On November 20, 2026, GitLab released GitLab CE and EE 19.6.1, 19.5.8 and 19.4.10. GitLab says that:
- These releases fix CVE-2026-132825 (GHSA-34m4-m9fh-gqp5), a heap-based buffer overflow in the case conversion of strings. Under some conditions, changing the case of a crafted string in GitLab can write outside a heap buffer.
- They also fix the other flaws in the Ruby advisory.
- The severity is Critical, from the Ruby advisory. The CVE record gives CVSS 3.1 7.4, High.
- GitLab.com and GitLab Dedicated already run a fixed version. Their customers need no action.
- The releases have no new migrations. A multi-node installation should not need downtime.
Upgrade each self-managed GitLab installation to 19.6.1, 19.5.8 or 19.4.10. GitLab gives no interim step.
Does this apply to you?
You have more exposure if one or more of these statements is true:
- You run self-managed GitLab CE or EE before 19.4.10, 19.5 before 19.5.8, or 19.6 before 19.6.1. This includes the Linux package, the Docker image and the Helm chart.
- You run a self-compiled GitLab with Ruby before 3.3.14, or Ruby 3.4 before 3.4.13. Upgrading GitLab does not change this Ruby.
- You run other Ruby applications on Ruby before 3.3.14, 3.4.0 to 3.4.12, or 4.0.0 to 4.0.9.
GitLab.com and GitLab Dedicated customers are not affected and need no action. GitLab says that GitLab Dedicated was protected earlier by internal releases.
Fixed releases
Your GitLab Linux package or Docker image Helm chart GitLab Operator 19.6 19.6.1 10.6.1 3.6.1 19.5 19.5.8 10.5.9 3.6.1 19.4 and earlier 19.4.10 10.4.11 3.6.1 The Helm chart versions are from the GitLab chart repository, which lists the GitLab release that each chart deploys. The GitLab Operator accepts only the charts in its own list. Operator 3.6.1 is the first release that lists all three fixed charts. Operator 3.6.0 and 3.5.6 do not. A release before 19.4 has no fix in its own branch: upgrade to a supported branch.
Recommended actions, in order
- Find your GitLab and Ruby versions. Run the checks on each GitLab node. Compare the result with the fixed releases table.
- Schedule the upgrade. GitLab says to upgrade immediately. Upgrade to the fixed release in your branch. With GitLab Geo, upgrade every site: the primary first, then each secondary.
- Plan for the restart. By default, the Linux package stops GitLab, runs migrations and starts it again during the upgrade. GitLab says these releases have no new migrations.
- On OpenShift, upgrade the GitLab Operator first. Then change the chart version in the GitLab resource. See "Upgrade GitLab on OpenShift".
- Rebuild Ruby on a self-compiled installation. Install Ruby 3.3.14 or 3.4.13, or a later release in the same series, then restart GitLab.
- Run the checks again after the upgrade. The Ruby version shows that the fix is in.
- Upgrade your other Ruby applications. Install Ruby 4.0.10, 3.4.13 or 3.3.14.
Check your GitLab and Ruby versions
These checks read the GitLab release and the Ruby that GitLab uses. They do not change the installation.
Linux package
Run these commands as root on each GitLab node.
shell cat /opt/gitlab/embedded/service/gitlab-rails/VERSION; echo /opt/gitlab/embedded/bin/ruby -vThe first line is the GitLab release. The second line is the Ruby that GitLab uses. The
VERSIONfile has no line end, soechoadds one. On GitLab 19.6.0, which is affected, the output is:text 19.6.0 ruby 3.3.13 (2026-10-23 revision af48f08f21) [x86_64-linux]On GitLab 19.6.1, which is fixed, the output is:
text 19.6.1 ruby 3.3.14 (2026-11-19 revision f4c37f481b) [x86_64-linux]GitLab 19.5.8 also shows
ruby 3.3.14. We ran these commands in the official GitLab CE 19.6.0 and 19.6.1 images, on x86_64 and on arm64 (aarch64-linux), and in the 19.5.8 image on arm64. We did not test 19.4.10 or GitLab EE.Docker
Run these commands on the Docker host. Replace
gitlabwith the name of your GitLab container.shell docker exec gitlab cat /opt/gitlab/embedded/service/gitlab-rails/VERSION; echo docker exec gitlab /opt/gitlab/embedded/bin/ruby -vThe output is the same as for the Linux package.
OpenShift and Kubernetes
Run these commands where you run
oc. Replacegitlabin-n gitlabwith the namespace of your GitLab resource. The pod names assume a GitLab resource or Helm release namedgitlab. On Kubernetes, usekubectlin place ofoc.shell oc -n gitlab exec deploy/gitlab-webservice-default -c webservice -- ruby -v oc -n gitlab exec deploy/gitlab-toolbox -c toolbox -- ruby -vOn a fixed release, each line starts with
ruby 3.3.14 (2026-11-19 revision f4c37f481b). Before the upgrade, our 19.5.5 test installation showedruby 3.3.12 (2026-07-16 revision 0581089df9). We also ranruby -vin the GitLab CE 19.6.1 webservice and toolbox images, and got the 3.3.14 line.Self-compiled installation
Run this command on the GitLab server. It shows the Ruby that the
gituser runs. We did not test this check.shell sudo -u git -H ruby -vThe result must be Ruby 3.3.14 or later in the 3.3 series, or 3.4.13 or later in the 3.4 series.
A higher version number alone is not enough. Ruby 3.4.0 to 3.4.12 are later than 3.3.13, but they are not fixed. On November 20, 2026, the GitLab page for self-compiled installations still named the earlier Ruby fixes, 3.3.13 and 3.4.12. Use the versions in the GitLab release note: 3.3.14 or 3.4.13.
Upgrade GitLab on OpenShift
On OpenShift, the GitLab Operator installs GitLab from the GitLab Helm chart. The Operator accepts only the chart versions in its own list, and the fixed charts need Operator 3.6.1. Upgrade the Operator first, then change the chart version in the GitLab resource at once. These steps are for an Operator that you installed with Helm.
Check the Operator settings
Operator chart 3.6.x renames the setting
extraEnvstoextraEnv. Look at your Operator settings before you upgrade:shell helm -n gitlab get values gitlab-operatorIf the output has
extraEnvs, rename it toextraEnvin your values.Upgrade the Operator
shell helm repo update helm -n gitlab upgrade gitlab-operator gitlab/gitlab-operator --version 3.6.1 --reuse-values --wait --timeout 10mChange the chart version
Run this command directly after the Operator upgrade. Use the chart for your branch from the fixed releases table:
10.6.1,10.5.9or10.4.11. If your GitLab resource also setsglobal.gitlabVersionor image tags, change them in the same patch.shell oc -n gitlab patch gitlab gitlab --type merge -p '{"spec":{"chart":{"version":"10.5.9"}}}' oc -n gitlab wait gitlab/gitlab --for=jsonpath='{.status.phase}'=Running --timeout=1800sThen run the OpenShift check in "Check your GitLab and Ruby versions".
Upgrade the Operator before you change the chart version. The old Operator refuses the new chart. In our test, Operator 3.5.5 refused chart 10.5.9 with
chart version 10.5.9 not supported; please use one of the following: 10.5.6, 10.4.8, 10.3.10.What we tested
We upgraded GitLab EE on OpenShift 4.19 from 19.5.5 (chart 10.5.6, Operator 3.5.5) to 19.5.8 (chart 10.5.9, Operator 3.6.1). We installed the Operator chart from its archive, not from the Helm repository. The results:
- The Operator upgrade took less than one minute. The GitLab resource was
Running10 to 13 minutes after the chart change. - The webservice, Sidekiq and toolbox containers changed from Ruby 3.3.12 to 3.3.14.
- Web and API reads did not fail. Our test installation has one Gitaly pod. While it restarted, for 43 seconds, two Git writes failed with HTTP 500. With more than one Gitaly replica, expect fewer failures.
We did not test the 19.6.1 or 19.4.10 upgrade on OpenShift, or an Operator that you installed from OperatorHub.
- The Operator upgrade took less than one minute. The GitLab resource was
Upgrade GitLab Geo sites
Each Geo secondary site runs the same Ruby as the primary, and serves web and Git traffic. An installation is fixed only when every site runs a fixed release.
Order
GitLab says to upgrade the primary site first, then each secondary site. On a zero-downtime upgrade, GitLab says to run the post-deployment migrations on the primary after all secondaries are upgraded.
On the Linux package, GitLab says to restart Puma and Sidekiq on each secondary after its upgrade. They then use the database schema that the secondary replicated from the primary. Run these commands on each secondary node:
shell sudo gitlab-ctl restart sidekiq sudo gitlab-ctl restart pumaOn OpenShift, upgrade the Operator and then change the chart version on the primary site, then do the same on each secondary site.
Check each site
Run the Geo check on an application node of the primary site and of each secondary site. GitLab says that all checks must pass:
shell sudo gitlab-rake gitlab:geo:checkThen run the Ruby check in "Check your GitLab and Ruby versions" on each site. On OpenShift, run the Geo check in the toolbox container:
shell oc -n gitlab exec deploy/gitlab-toolbox -c toolbox -- gitlab-rake gitlab:geo:checkRead the output, not only the exit code. The check can exit with code 0 when a line says
no. On a secondary site, the lines for the tracking database, replication, the connection to the primary, the node name and the clock must sayyes. The primary site shows a shorter list.On OpenShift, one
noline is expected. The line "OpenSSH configured to use AuthorizedKeysCommand" saysno, because the toolbox container has no SSH server. In our test it saidnoon both sites, and Geo was Healthy.What we tested
In our OpenShift 4.19 test, with one primary and one secondary site:
- The secondary ran 19.5.5 against the 19.5.8 primary for approximately 16 minutes. Geo stayed Healthy, with no replication lag.
- No Git write that the primary accepted was missing on the secondary after the upgrade.
- On a secondary that pulls images from a local mirror, nine GitLab images changed to tag
v19.5.8, and the Operator image changed to3.6.1. The container registry and GitLab exporter images did not change. Copy the images to the mirror before you upgrade the primary. Add the images of the components that you enable.
We ran the OpenShift Geo check on both sites after the upgrade. We did not test Geo on the Linux package.
What to monitor
- GitLab patch releases. GitLab releases scheduled patch releases on the second and fourth Wednesday of each month, and critical patch releases when necessary. Subscribe to the GitLab security notices.
- Crashes of GitLab Ruby processes. These flaws corrupt memory. A crash of Puma or Sidekiq before you upgrade can be a sign of an attack. This is our analysis, not a GitLab statement.
- The CVE records. The Ruby advisory uses GHSA identifiers, because CVE IDs were not assigned on November 18, 2026. Watch for the CVE IDs of the other six flaws.
References
- GitLab, Critical Patch Release: 19.6.1, 19.5.8, 19.4.10 (opens in a new tab): the fixed releases, the affected versions, the severity, the GitLab.com and GitLab Dedicated status, and the guidance for self-compiled installations.
- Ruby, multiple vulnerabilities in Regexp and String case conversion (opens in a new tab): the seven flaws, the affected and fixed Ruby versions, and the proof-of-concept code.
- GitHub, GHSA-34m4-m9fh-gqp5 (opens in a new tab): the case conversion flaw, the encodings and the fixed Ruby versions.
- CVE, CVE-2026-132825 (opens in a new tab): the CVE record and its CVSS score.
- GitLab, self-compiled installation (opens in a new tab): the Ruby step for self-compiled installations.
- GitLab, upgrading the Geo sites (opens in a new tab): the order of the sites, the Puma and Sidekiq restart and the Geo check.
- GitLab, zero-downtime upgrades (opens in a new tab): the Geo order and the post-deployment migrations.
- GitLab Operator, the
CHART_VERSIONSfile at tags 3.5.6, 3.6.0 and 3.6.1: the charts that each Operator release accepts. - GitLab Operator chart 3.6.1: the
extraEnvsetting. - GitLab Helm chart repository: the chart version for each GitLab release.
About this advisory
This page gives the results of our analysis. It is not a statement from GitLab or from the Ruby security team. GitLab has confirmed the fixed releases, the affected versions, the severity and the status of GitLab.com and GitLab Dedicated. The Ruby security team has confirmed the flaws and the fixed Ruby versions. The checks, the Helm chart and Operator versions, the OpenShift and Geo steps and the monitoring signs are our work. We tested the Linux package and Docker checks in the official GitLab CE 19.6.0, 19.6.1 and 19.5.8 images. We did not test the self-compiled check, GitLab EE or the Helm chart. The code is licensed under the Apache License 2.0 and is supplied as it is, with no warranty. You are responsible for tests in your environment. We will update this page when GitLab changes its release note, and when the other Ruby flaws get CVE IDs.