The system is a Sun U5 running Solaris 2.6 and Squid 2.2S5; it has 512 MB
RAM and is doing approximately 25 reqs / second.  squid is built with
asyncio turned on.  sar -r does not indicate any significant drop in
available system memory:
11:45:00    1234  1947742
12:00:01    1280  1949016
12:15:00    1177  1947806
12:30:01    1166  1948487
12:45:01    1251  1947080
13:00:01    1122  1945463
13:15:06    1250  1943471
13:30:00    1158  1946548
13:45:09    1292  1945131
14:00:01    1171  1944605
14:15:01    1148  1946769
14:30:01    1167  1945919
14:45:00    1230  1944514
15:00:01    1251  1947210
15:15:00    1260  1945546
15:30:00    1296  1946010
15:45:00    1206  1847972
16:00:01    3166  1759425
This machine, however, produced the following error around 15:40:
libthread panic: _sys_thread_create():alloc_thread returns 0 (no mem) (PID:
1642
1 LWP 1)
stacktrace:
        ef5d3140
        ef54ff28
        ef54fa80
        31f34
        2e740
        3108c
        312ac
        31a48
        4d27c
        1aab4
        0
Looking through the FAQ, there was mention of a memory leak in dlmalloc so I
tried to build using Sun's native malloc; this seemed to exacerbate the
problem - the errors were coming fast and furious, probably every two or
three hours.  It took approximately three days for this problem to recur
with dlmalloc compiled back in.
  Any ideas from the group?
  
  Thanks.
________________________________________________________________
Get FREE voicemail, fax and email at http://voicemail.excite.com
Talk online at http://voicechat.excite.com
Received on Fri Oct 08 1999 - 10:48:15 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:48:45 MST