You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
shows no platforms specified in versions.json as being dropped in the default main branch of node, so it should not be necessary to drop any platforms in versions.json for the Docker builds.
Debian releases shows Debian 13 (trixie) as the current "Stable" release with End of LTS on 2030-06-30, which is after the planned EOL of Node.js 26 on 2029-04-30. Debian 12 (bookworm) End of LTS 2028-06-30 & Debian 11 (bullseye) End of LTS 2026-08-31 dates are earlier than EOL of Node.js 26.
Alpine Linux shows v3.21 - v3.23 as supported when Node.js is released.
Suggestions
For the upcoming Node.js 26 release:
Change default from Debian 12 (bookworm) to Debian 13 (trixie)
Item
25
26
debian-default
bookworm
trixie
Drop oldoldstable & oldstable Debian builds
I suggest to drop builds for Debian 11 (bullseye / bullseye-slim / oldoldstable) & Debian 12 (bookworm / bookworm-slim / oldstable) for Node.js 26 and build only for Debian 13 (trixie / stable), unless there is a compelling reason to continue with the oldoldstable & oldstable releases. This significantly reduces the number of build combinations. See also Debian future release strategy proposal #2398 regarding defining a strategy for Debian releases.
For Alpine, continue with v3.22 & v3.23 and with alpine-default as alpine3.23
Is it desirable to drop Debian oldoldstable & oldstable for the upcoming Node.js 26 release?
Are the current set of scripts supposed to support a setup where Node.js 26 would have a different set of variants compared to previously Node.js releases? I originally assumed that versions / variants would be completely data-driven from versions.json, however it appears that is not the whole story!
Only support trixie for 26 sounds good to me but it would be good to hear if others are aware of any particular precedent/policy in this area.
ping @nodejs/docker
I'd keep all the variants, just so that people pulling the variant tags keep getting the "current" release, and don't accidentally end up with a 25 image that stops being patched. When the variants hit EOL, we've previously removed them from the matrix, even if the Node version is still supported.
I agree with setting the default variant up to trixie to start the release line. I don't recall if that might confuse the stackbrew PR generation though.
For most of the Docker Official Images that I directly maintain, I try to keep it to just two OS versions per OS, for example Trixie and Bookworm along with alpine 3.23 and 3.22. But it is really up to the node team how many OS versions that you want to support (until their respective end of life dates).
I usually point floating tags that don't specify an OS release to the latest OS; for example, 26 -> Trixie (and would be Forky when it's released), 26-alpine -> Alpine 3.23 (and 3.24 when released). This forces users to either pin to an OS version (and then know that they need to consider when to upgrade) or to actively grapple with OS changes when it moves.
It seems that the consensus and short-term viable way forward for the Node.js 26.0.0 release this week would be according to #2435 (comment) from @nschonni above
keep all variant tags
set default to trixie
The strategy suggestion from @yosifkit to restrict to 2 major OS releases is definitely attractive as a long-term strategy. This fits also what GitHub Actions does with their runner images.
@nschonni I'll note that Bookworm's LTS schedule covers up to the end of June 2028 which is likely to be before the end of Node 26 which would be a reason to do Trixie only as we may have to switch off bookworm during the Node 26 LTS cycle which IMHO isn't ideal although as you mentioned we've historically done exactly that. When Debian Forky comes out net year we might choose to support that in the Node 26 cycle which would bump us up to 3, unless we decided to ditch bookworm at that point ... i'm not sure if we currently have an explicit policy of when we have introduced new Debian verisons.
Having said that, i'm ok with leaving it in for now and we can determine a more official policy for Node 27 next year if that's the consensus.
EDIT: Also ref doc issue on this topic at #2397 and Debian specifically at #2398
I'm closing this issue and suggest that any other discussion is carried out directly in PR #2474 (which looks fine to me, apart from the missing hashes that won't be available until the actual release is available).
Activity
MikeMcC399 commentedon Mar 31, 2026
Questions for maintainers:
oldoldstable&oldstablefor the upcoming Node.js 26 release?sxa commentedon Apr 8, 2026
Only support trixie for 26 sounds good to me but it would be good to hear if others are aware of any particular precedent/policy in this area.
ping @nodejs/docker
nschonni commentedon Apr 8, 2026
I'd keep all the variants, just so that people pulling the variant tags keep getting the "current" release, and don't accidentally end up with a 25 image that stops being patched. When the variants hit EOL, we've previously removed them from the matrix, even if the Node version is still supported.
I agree with setting the default variant up to
trixieto start the release line. I don't recall if that might confuse thestackbrewPR generation though.yosifkit commentedon Apr 8, 2026
For most of the Docker Official Images that I directly maintain, I try to keep it to just two OS versions per OS, for example Trixie and Bookworm along with alpine 3.23 and 3.22. But it is really up to the node team how many OS versions that you want to support (until their respective end of life dates).
I usually point floating tags that don't specify an OS release to the latest OS; for example,
26-> Trixie (and would be Forky when it's released),26-alpine-> Alpine 3.23 (and 3.24 when released). This forces users to either pin to an OS version (and then know that they need to consider when to upgrade) or to actively grapple with OS changes when it moves.sxa commentedon Apr 20, 2026
Adding link to the PR which added 25 support for reference: #2293
MikeMcC399 commentedon Apr 20, 2026
It seems that the consensus and short-term viable way forward for the Node.js 26.0.0 release this week would be according to #2435 (comment) from @nschonni above
The strategy suggestion from @yosifkit to restrict to 2 major OS releases is definitely attractive as a long-term strategy. This fits also what GitHub Actions does with their runner images.
Who is going to submit the PR to add Node.js 26?
The one for Node.js 25 (#2293) was done by @PeterDaveHello
sxa commentedon Apr 20, 2026
@nschonni I'll note that Bookworm's LTS schedule covers up to the end of June 2028 which is likely to be before the end of Node 26 which would be a reason to do Trixie only as we may have to switch off bookworm during the Node 26 LTS cycle which IMHO isn't ideal although as you mentioned we've historically done exactly that. When Debian Forky comes out net year we might choose to support that in the Node 26 cycle which would bump us up to 3, unless we decided to ditch bookworm at that point ... i'm not sure if we currently have an explicit policy of when we have introduced new Debian verisons.
Having said that, i'm ok with leaving it in for now and we can determine a more official policy for Node 27 next year if that's the consensus.
EDIT: Also ref doc issue on this topic at #2397 and Debian specifically at #2398
[-]Prepare for Node.js 26 release 2026-04-22[/-][+]Prepare for Node.js 26 release Apr 2026[/+]MikeMcC399 commentedon Apr 20, 2026
According to nodejs/node#62526 (comment) the v26 release is delayed.
MikeMcC399 commentedon Apr 21, 2026
Since this issue was only about preparing for Node.js 26.0.0 and not about actually releasing it, and
I'm closing this issue and suggest that any other discussion is carried out directly in PR #2474 (which looks fine to me, apart from the missing hashes that won't be available until the actual release is available).