-
Notifications
You must be signed in to change notification settings - Fork 759
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Jail cleanup issue. #9976
Comments
[pid 873027] 1725454042.765797 stat("/opt/cool/child-roots/873022-20fbac79/KU7MORiIpumls3Mj/lo", 0x7ffccb26be68) = -1 ENOENT (No such file or directory) To take an example: ls -ld QJ* So - that should be cleaned; it contains: ls -alR QJ90EiwQaou8PdUX/ Nothing =) |
So the check happens every 5 seconds in the forkit loop - surprising to see it as significant on the profile; but I guess ... 12 million stats per day for the 700 files is something (if I count right). Not a hyper-high-priority, but would be nice to get cleaned up. |
They used namespace mounts for / so that was both unmounted automatically when the child exited, and not visible to the parent anyway, so these should be empty, with no "lo" to check for, or any other content. #9976 Signed-off-by: Caolán McNamara <[email protected]> Change-Id: Ia98032abaae8a33abd57cef38bdc97f6ff4fb0c2
which is where path/<jailid>/tmp is mounted (or linked) from. This seems to have been the case before namespace support, but it is just an empty dir without namespaces, but with namespaces (because the path/<jailid>/tmp mount is hidden from parent processes) the jail tmp contents were also left behind. #9976 Signed-off-by: Caolán McNamara <[email protected]> Change-Id: If4aed3ffa0eca73e915f6d48c50a7e70601e210d
They used namespace mounts for / so that was both unmounted automatically when the child exited, and not visible to the parent anyway, so these should be empty, with no "lo" to check for, or any other content. #9976 Signed-off-by: Caolán McNamara <[email protected]> Change-Id: Ia98032abaae8a33abd57cef38bdc97f6ff4fb0c2
which is where path/<jailid>/tmp is mounted (or linked) from. This seems to have been the case before namespace support, but it is just an empty dir without namespaces, but with namespaces (because the path/<jailid>/tmp mount is hidden from parent processes) the jail tmp contents were also left behind. #9976 Signed-off-by: Caolán McNamara <[email protected]> Change-Id: If4aed3ffa0eca73e915f6d48c50a7e70601e210d
They used namespace mounts for / so that was both unmounted automatically when the child exited, and not visible to the parent anyway, so these should be empty, with no "lo" to check for, or any other content. CollaboraOnline/online#9976 Adapt patch to LibreOffice Online. Signed-off-by: Caolán McNamara <[email protected]> Change-Id: Ia98032abaae8a33abd57cef38bdc97f6ff4fb0c2
which is where path/<jailid>/tmp is mounted (or linked) from. This seems to have been the case before namespace support, but it is just an empty dir without namespaces, but with namespaces (because the path/<jailid>/tmp mount is hidden from parent processes) the jail tmp contents were also left behind. CollaboraOnline/online#9976 Signed-off-by: Caolán McNamara <[email protected]> Change-Id: If4aed3ffa0eca73e915f6d48c50a7e70601e210d
Share profile as of now:
6% of server time in that statting - which seems amazing; perhaps there is some explosion of crashing kits or something - is it possible that: if (time(nullptr) < modifiedTimeSec + 180)
continue;
doens't work - and we build a huge list of crashed kits - but the; kits should not crash on share much - a tad perplexed by that =)
ls /opt/cool/child-roots/873022-20fbac79/ | wc -l
742
interesting.
vs. 64 coolforkit processes
It looks like we are leaking jails.
The text was updated successfully, but these errors were encountered: