We are having a problem with OSF 1 on a DEC Alpha 3000/300. Uname -a
on that machine says:
OSF1 caesar.ultra.net V3.0 358.78 alpha
The problem is symptomized by the refusal of our 'nnrpd' task (a
daemon from a publicly-available package which provides service to
clients using the NNTP (network news transport) protocol) to accept
connections from clients. The symptom occurs at our peak news reader
usage time, between about 8:30 p.m. and about 10:00 p.m.
I have asked other users of the INN package (from which nnrpd derives)
and nobody else has ever seen this particular problem, albeit nobody I
asked is running the package under OSF.
The following syslog messages appear to coincide with each instance of
server refusal, and it is these messages that we are seeking better to
understand:
Jun 8 21:32:11 caesar vmunix: fork/procdup: task_create failed. Code: 0x11
Jun 8 21:32:37 caesar last message repeated 8 times
Jun 8 21:34:56 caesar last message repeated 6 times
Jun 8 21:35:03 caesar last message repeated 3 times
According to the man page, fork() should return an errno, but this
particular one doesn't seem to make any sense: 0x11 = 17 = EEXIST =
"file exists."
Our main need is to understand what this code means, and to understand
why it's happening. We want to understand what resource it's starved
for so that we can take corrective action.
We have been compiling various system statistics on a 15-minute
frequency, and happened to take such a snapshot at 21:35, which is
around the time the problem was occuring. Here are the salient
portions of the stats we collected, from 21:20, and 21:35, and 21:50:
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Thu Jun 8 21:20:04 EDT 1995
caesar.ultra.net
21:20 up 13 days, 15:26, 0 users, load average: 1.29, 1.08, 1.48
========== News server tasks: 42
Total swap allocation:
Allocated space: 49152 pages (384MB)
Reserved space: 9088 pages ( 18%)
In-use space: 7799 pages ( 15%)
Available space: 40064 pages ( 81%)
Virtual Memory Statistics: (pagesize = 8192)
procs memory pages intr cpu
r w u act free wire fault cow zero react pin pout in sy cs us sy id
3 68 17 4493 117 2222 22M 1M 9M 1M 3M 858K 415 537 349 42 33 26
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Thu Jun 8 21:35:16 EDT 1995
caesar.ultra.net
21:35 up 13 days, 15:41, 0 users, load average: 10.89, 9.81, 9.11
========== News server tasks: 55
Total swap allocation:
Allocated space: 49152 pages (384MB)
Reserved space: 10705 pages ( 21%)
In-use space: 9305 pages ( 18%)
Available space: 38447 pages ( 78%)
Virtual Memory Statistics: (pagesize = 8192)
procs memory pages intr cpu
r w u act free wire fault cow zero react pin pout in sy cs us sy id
10 69 20 4422 97 2251 22M 1M 9M 1M 3M 860K 416 538 350 42 33 26
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Thu Jun 8 21:50:04 EDT 1995
caesar.ultra.net
21:50 up 13 days, 15:56, 0 users, load average: 3.69, 3.12, 2.83
========== News server tasks: 43
Total swap allocation:
Allocated space: 49152 pages (384MB)
Reserved space: 9406 pages ( 19%)
In-use space: 8099 pages ( 16%)
Available space: 39746 pages ( 80%)
Virtual Memory Statistics: (pagesize = 8192)
procs memory pages intr cpu
r w u act free wire fault cow zero react pin pout in sy cs us sy id
4 62 23 4494 126 2218 22M 1M 9M 1M 3M 861K 416 538 350 42 33 26
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
We look forward to hearing any helpful comments that readers who have
some internal knowledge of OSF might be able to offer.
Thank you.
Dwight A. Ernest
Network and Systems Administration Team, UltraNet Communications Inc.
508.229.8400(voice) noc_at_ultra.net 508.229.8111(data)
A network service provider in Central Mass mailto:info_at_ultra.net
Received on Fri Jun 09 1995 - 21:21:23 NZST