[squid-users] xmalloc: Unable to allocate 87380 bytes!

Over the last couple of days, Squid has stopped unexpectedly with the
error "xmalloc: Unable to allocate 87380 bytes!". I have included a
portion of the relevant cache.log below as well as the output of "vmstat
1" during that time.

Sep 30 12:54:58 proxy squid[6365]: xmalloc: Unable to allocate 87380 bytes!
Sep 30 12:54:58 proxy squid[6365]: storeDirWriteCleanLogs: Starting...
Sep 30 12:54:58 proxy squid[6365]: WARNING: Closing open FD 23
Sep 30 12:54:58 proxy squid[6365]: WARNING: Closing open FD 45
Sep 30 12:54:58 proxy squid[6365]: WARNING: Closing open FD 46
Sep 30 12:54:58 proxy squid[6365]: WARNING: Closing open FD 47
Sep 30 12:54:58 proxy squid[6365]: 65536 entries written so far.
Sep 30 12:54:58 proxy squid[6365]: 131072 entries written so far.
Sep 30 12:54:58 proxy squid[6365]: 196608 entries written so far.

   procs memory swap io system
 cpu
 r b w swpd free buff cache si so bi bo in cs us
sy id
 1 0 1 0 11632 176484 786268 0 0 0 15580 1748 2265 6
9 85
 0 2 2 0 11616 176184 786676 0 0 0 16800 1780 2571 6
7 86
 1 0 1 0 11396 176224 786616 0 0 0 16080 1934 2892 10
4 85
 0 1 1 0 10816 174112 788848 0 0 0 16688 1741 2313 6
6 88
 1 1 1 0 11632 174212 788924 0 0 0 16284 1114 470 14
14 72
 2 1 1 0 11512 171008 792380 0 0 4 26928 1005 1516 33
21 47
 0 1 1 0 291484 170664 789400 0 0 352 14412 935 474 0
11 89
 0 0 0 0 952812 170708 789444 0 0 88 12288 3204 134 0
14 86
 0 0 0 0 952804 170716 789444 0 0 8 0 418 70 0
0 100
 0 0 1 0 951316 172204 789444 0 0 0 4680 1101 238 0
1 99
 0 0 0 0 951308 172212 789444 0 0 0 84 345 81 0
0 100
 0 0 0 0 951308 172212 789444 0 0 0 0 320 83 0
0 100

This is a somewhat new system, only been in use for a couple of months.
The cache directories just reached capacity within the last couple of
days. The Squid version is 2.4s7. The OS is Redhat 7.3 running on:

Dell Poweredge 2500
Dual 1.4 GHz P3
2 GB Ram
(6) 18 GB 15K SCSI Disk Drives
Hardware Raid Controller

The disks are partitioned as:
  2 GB of each disk as raid 5 for the operating system
  14.5 GB of each disk JBOD for the cache directories

Squid compile time options:

Prefix = /usr/local/squid
Enable-dlmalloc
Enable-gnuregex
Enable-async-io
Enable-removal-policies=heap,lru
Enable-underscores
Enable-storeio=aufs,ufs
Disable-ident-lookups
Enable-snmp

Our previous system never had this problem. It was running an older
version of Redhat (can't remember which one exactly), single processor, 1
GB ram.

I'm thinking that the problem might be a kernel VM issue (kernel is redhat
standard 2.4.18-10smp), but I couldn't find any tunables that seemed like
they would help.

Any help fixing this would be appreciated.

Thanks,

Erik
Received on Mon Sep 30 2002 - 14:33:22 MDT