livelock without explaination.

Zaphod Beeblebrox zbeeble at gmail.com
Thu Oct 20 20:58:42 UTC 2016


I have a FreeBSD 10.3 server that houses a 16 disk ZFS array, runs samba,
and runs a large torrent client (a few hundred torrents, but probably less
than 50 of them active).

Every so often it goes into a livelock.  The pipe from the location is
50/10 ... and the torrents are limited to 15/5.

Inside the building, GigE access through SMB or NFS isn't uncommon, but not
terribly constant.  Watching movies happens over SMB, but I hardly think
that's a stretch.

... but every 3 to 5 days, I get livelock.  Even the console locks,
although ... often... after some minutes, I can get a CTRL-ALT-DEL to
register and get it shut down properly.  It's also true that stopping
rtorrent will unfreeze the livelock, but it will return quickly if
restarted.

I've  tried reducing the number of active torrents.  I've also looked at
the netstat -m

430/15965/16395 mbufs in use (current/cache/total)
21/2065/2086/2034610 mbuf clusters in use (current/cache/total/max)
21/2003 mbuf+clusters out of packet secondary zone in use (current/cache)
128/249/377/1017304 4k (page size) jumbo clusters in use
(current/cache/total/max)
256/2775/3031/301423 9k jumbo clusters in use (current/cache/total/max)
0/0/0/169550 16k jumbo clusters in use (current/cache/total/max)
2965K/34092K/37057K bytes allocated to network (current/cache/total)
0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)
0/0/0 requests for mbufs delayed (mbufs/clusters/mbuf+clusters)
0/0/0 requests for jumbo clusters delayed (4k/9k/16k)
0/0/0 requests for jumbo clusters denied (4k/9k/16k)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile

... now I'd like to increase mbufs there ... but:

[1:5:305]root at vr:~> sysctl -a | grep nmb
kern.ipc.nmbufs: 13021500
kern.ipc.nmbjumbo16: 678200
kern.ipc.nmbjumbo9: 904269
kern.ipc.nmbjumbop: 1017304
kern.ipc.nmbclusters: 2034610

... so... odd... and why would this all effect typing on the keyboard ...
unless there's some interaction between the disk controllers and this lot?
The root is UFS on a SATA gmirrored with a PATA drive (legacy).  This that
really bad (tm)?


More information about the freebsd-hackers mailing list