ports/131554: flow-capture consuming all available memory

[Oleg Gawriloff]

>Number:         131554
>Category:       ports
>Synopsis:       flow-capture consuming all available memory
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    freebsd-ports-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Tue Feb 10 09:10:02 UTC 2009
>Closed-Date:
>Last-Modified:
>Originator:     Oleg Gawriloff
>Release:        7.1
>Organization:
Atlant Telecom
>Environment:
FreeBSD netflow.telecom.by 7.1-RELEASE-p2 FreeBSD 7.1-RELEASE-p2 #0: Tue Jan 27 08:59:57 EET 2009     root at netflow.telecom.by:/usr/obj/usr/src/sys/NETFLOW  amd64

>Description:
We have heavily loaded netflow server using flow-tools/flow-extract. During time (3-5 days) flow-extract consuming all available memory on server (i.e. 6Gb) and then server stops responding with "+swap_pager_getswapspace(16): failed". Now we solved this problem using monit with flow-extract restart when its memory consumption above 3Gb, but may be there is another way.

Typical top output after 1 day of work:
last pid: 98873;  load averages:  0.95,  1.17,  1.11                                          up 10+12:15:37  11:08:15
115 processes: 13 running, 86 sleeping, 16 waiting
CPU: 14.5% user,  0.0% nice,  1.4% system,  0.0% interrupt, 84.0% idle
Mem: 898M Active, 2947M Inact, 347M Wired, 1236K Cache, 214M Buf, 742M Free
Swap: 2048M Total, 13M Used, 2035M Free

  PID USERNAME     THR PRI NICE   SIZE    RES STATE  C   TIME    CPU COMMAND
68525 root           1  52    0   844M   805M CPU0   0  46:21 14.26% flow-capture

>How-To-Repeat:

>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted: