Robin Humble wrote:
now onto the next problem :)
setup is a single ~1TB software raid0 OST on 'xe', an MDT/MGS on 'x19'
with SATA disk, and 16 other client nodes mounting the Lustre filesystem
and doing bonnie++ over o2ib. all standard Lustre rpms, x86_64 CentOS4.4.
IIRC the variable number of threads is a new feature in 1.5.97?
looks like there's a small bug in it...
it ends up with 512 ll_ost_io_* threads, and the last is hung in a D state
with no i/o on the filesystem. ie.
% ps auxw
...
root 7488 0.0 0.0 0 0 ? S 19:20 0:00 [ll_ost_io_511]
root 7489 0.0 0.0 0 0 ? D 19:20 0:00 [ll_ost_io_512]
...
after the i/o:
num threads name
----------- ----
5 ll_log_comt_*
8 ldlm_bl_*
8 ldlm_cb_*
15 ldlm_cn_*
304 ll_ost_*
512 ll_ost_io_*
the /tmp/lustre-log.* files are kinda huge. they're at
http://www.cita.utoronto.ca/~rjh/lustre/
there's no LBUG if I restrict the number of threads with eg:
options ost oss_num_threads=300
let me know if you'd like more info or to try patches etc.
cheers,
robin
Feb 10 19:20:41 xe kernel: LustreError: 7489:0:(ost_handler.c:1555:ost_thread_init())
ASSERTION(thread->t_id < OSS_THREADS_MAX) failed
Looks like it's just a slightly bad assert.
Index: ost_handler.c
===================================================================
RCS file: /cvsroot/cfs/lustre-core/ost/ost_handler.c,v
retrieving revision 1.181
diff -u -p -r1.181 ost_handler.c
--- ost_handler.c 10 Feb 2007 06:32:49 -0000 1.181
+++ ost_handler.c 13 Feb 2007 20:54:19 -0000
@@ -1586,7 +1586,7 @@ static int ost_thread_init(struct ptlrpc
LASSERT(thread != NULL);
LASSERT(thread->t_data == NULL);
- LASSERT(thread->t_id < OSS_THREADS_MAX);
+ LASSERT(thread->t_id <= OSS_THREADS_MAX);
OBD_ALLOC_PTR(tls);
if (tls != NULL) {
_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
