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
{{ message }}
Repository navigation
wasm32-wasi 3.x buildbot: No space left on device #154532
/usr/bin/ld.bfd: final link failed: No space left on device
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make: *** [Makefile:1791: _bootstrap_python] Error 1
@brettcannon: I'm sorry! Your previous attempt to free enough disk space on wasm32-wasi 3.x buildbot was not enough. The "No space left on device" error is back.
cc @stratakis: We do have frequent "No space left on device" in our Fedora/CentOS/RHEL buildbot workers as well :-(
There's plenty of space on that disk. That df output if after stopping podman running https://github.com/brettcannon/wasi-buildbot/blob/main/run.py and deleting the checkouts which are mounted in ~buildbot/builldarea which were like 23 GB, so still well under 52 GB. I can try checking when it happens again with df to see what's actually running out of disk space.
ar rcs libpython3.16d.a Modules/getbuildinfo.o (...) Python/frozen.o
ar: unable to copy file 'libpython3.16d.a'; reason: No space left on device
make: *** [Makefile:1164: libpython3.16d.a] Error 1
I've once again deleted the checkouts and the buildbot is starting up again, but is this amount of space usage expected? I thought some change was made a while back to use less space in the Git checkout?
The CPython repo has grown significantly. Waiting for a new release of buildbot for buildbot/buildbot#9058 to hopefully resolve some of the disk space issues.
Got an email that it's happened again, so when I have time I will clear out the space and then turn on shallow checkouts as the SC has not told me I can just skip having a buildbot yet.
Activity
brettcannon commentedon Jul 23, 2026
I'm running out of ideas here.
There's plenty of space on that disk. That
dfoutput if after stopping podman running https://github.com/brettcannon/wasi-buildbot/blob/main/run.py and deleting the checkouts which are mounted in~buildbot/builldareawhich were like 23 GB, so still well under 52 GB. I can try checking when it happens again withdfto see what's actually running out of disk space.vstinner commentedon Jul 24, 2026
The buildbot worker is back to green, so I close the issue.
vstinner commentedon Aug 14, 2026
I reopen the issue, "No space left on device" is back. cc @brettcannon
Build12739 failed with "Bus error": https://buildbot.python.org/#/builders/1046/builds/12739
The following build 12740 failed with "No space left on device": https://buildbot.python.org/#/builders/1046/builds/12740
brettcannon commentedon Aug 14, 2026
The checkouts are taking up over 50 GB:
Is that expected to be happening?
https://devguide.python.org/testing/new-buildbot-worker/#required-resources says I should only need about 30 GB of free space.
I've once again deleted the checkouts and the buildbot is starting up again, but is this amount of space usage expected? I thought some change was made a while back to use less space in the Git checkout?
brettcannon commentedon Aug 14, 2026
Zach pointed me at https://github.com/python/buildmaster-config/blob/d74e3cbef971ca5e5c1cfeb2a6740343592ccfa1/master/custom/workers.py#L341-L348 as a possible solution as they think this is due to a bug in buildbot itself.
stratakis commentedon Aug 15, 2026
The CPython repo has grown significantly. Waiting for a new release of buildbot for buildbot/buildbot#9058 to hopefully resolve some of the disk space issues.
vstinner commentedon Aug 31, 2026
https://buildbot.python.org/#/builders/1046 is back to green, I close the issue.
brettcannon commentedon Sep 2, 2026
Got an email that it's happened again, so when I have time I will clear out the space and then turn on shallow checkouts as the SC has not told me I can just skip having a buildbot yet.
brettcannon commentedon Sep 8, 2026
I've opened python/buildmaster-config#788 to use shallow clones to try and work around the disk usage problem.