winbindd stops and takes up 60%-100% CPU

Bug #282825 reported by Brandon Perry
12
This bug affects 1 person
Affects Status Importance Assigned to Milestone
samba (Ubuntu)
Confirmed
Undecided
Unassigned

Bug Description

After the winbind updates this weekend/this morning, winbind will lock up and start using 60%-100% of the CPU. It actually for some reason has two instances of winbindd running, each using between 60% and 100% of the CPU (two cores).

Distro:
Intrepid

Package details:
ii libwbclient0 2:3.2.3-1ubuntu3 client library for interfacing with winbind service
ii winbind 2:3.2.3-1ubuntu3 service to resolve user and group information from Wind

Kernel:
2.6.27-7-generic #1 SMP Fri Oct 10 03:55:24 UTC 2008 i686 GNU/Linux

Revision history for this message
Brandon Perry (bperry-volatile) wrote :

I would also like to add that because of winbindd locking up, I cannot use su(do) and therefore cannot kill or restart the winbindd daemon (my tty's don't work, so I can't go through a tty). I have to restart my computer in order to get everything working again, which lasts maybe 10-15 minutes until it locks up again.

Revision history for this message
Jelmer Jaarsma (jelmer-jaarsma) wrote :

I was having this problem as well, is your winbindd also segfaulting? It seems much related to the segfaults I reported over here: https://bugs.launchpad.net/ubuntu/+source/samba/+bug/282733

My "workaround" was issueing a "getent passwd", that seemed to wake winbindd up and make it work again

Revision history for this message
Aaron C. de Bruyn (darkpixel2k) wrote :
Download full text (4.3 KiB)

This issue is happening to me using the copy of winbind that comes with the likewise-open package.
root@crosstalk:/etc/samba# apt-cache policy likewise-open
likewise-open:
  Installed: 4.0.5-0ubuntu3.1
  Candidate: 4.0.5-0ubuntu3.1
  Version table:
 *** 4.0.5-0ubuntu3.1 0
        500 http://us.archive.ubuntu.com hardy-updates/main Packages
        100 /var/lib/dpkg/status
     4.0.5-0ubuntu3 0
        500 http://us.archive.ubuntu.com hardy/main Packages

This is on an intranet webserver, so it's mostly used to authenticate people hitting apache against MS Active Directory.
I usually see two winbind processes maxing out the CPU between them.

If I attach to either process using strace while this is going on, I get the following:
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=1093056, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=395548, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=395548, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=420456, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=420456, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=445364, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=445364, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=460800, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=460800, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=560112, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=560112, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=608776, len=1}, 0xbf8b9f74) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f84) = 0
fcntl64(11, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=5820, len=1}, 0xbf8b9f14) = 0
fcntl64(11, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=608776, len=1}, 0xbf8b9f64) = 0
fcntl64(11, F_SETLKW64, {type=F_RDLCK, when...

Read more...

Changed in samba:
status: New → Confirmed
Revision history for this message
Aaron C. de Bruyn (darkpixel2k) wrote :

Also, I can provide developers SSH access to the box if needed for debugging.

Revision history for this message
Chuck Short (zulcss) wrote :

Hi,

I was wondering if you still having this problem?

Regards
chuck

Revision history for this message
Aaron C. de Bruyn (darkpixel2k) wrote :

I am still running into the issue. I just upgraded the machine this morning from 8.10 to 9.04. In a few weeks it will be brought up to 9.10.

The issue is most noticable with apache because the machine serves our corporate intranet. Whenever people try to access pages it takes 3-5 seconds for the page to return, and the entire time the likewise-winbindd process is using up significant CPU time.

Revision history for this message
Tony Wood (twood) wrote :

Possible fix:
Greetings: We have been getting CPU 40-50% for months on winbind. I have finally got of my backside as had a go at looking for a fix. I have just done the following commands and the CPU dropped to 1% It basically re-joins the Domain, even though we already have a Server account in the AD

sudo kinit <email address hidden>
net ads join -U Administrator

Good luck
Tony

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.