Why don't you run it under gdb, let it serve as usual and wait for the
crash, then just issue a backtrace command. Or at least let it dump
core when it crashes.
Brandon Ooi wrote:
Hi,
Our memcache machines seem to be dying off randomly after a long time
and we can't figure out why. We're running them on 32bit FC 6 machines
with 4GB or ram.
There's no messages in /var/log/messages or dmesg.
Here's a "snapshot" of the stats of one of the memcache machines
before it died. Does anybody have any ideas as to why this could be or
does any of this look weird?
pid 2426
uptime 1155776
time 1180585163
version 1.2.2
pointer_size 32
rusage_user 7756.163883
rusage_system 23106.066343
curr_items 5681384
total_items 86312412
bytes 2.56 GB
curr_connections 1332
total_connections 2265133
connection_structures 5399
cmd_get 1095198349
cmd_set 86312412
get_hits 1024085078
get_misses 71113271
evictions 0
bytes_read 106.39 GB
bytes_written 896.24 GB
limit_maxbytes 3.42 GB
threads 1
'qps' 1022.27
hit/miss ratio 14.40
get/set ratio 12.69
write/read ratio 8.42
Brandon
