|
Hi,
I just did some syn-flooding tests to the ipvs box. The number of inactive
connections reach over 200,000 soon, but the system is almost idle.
[root@koala ipvs]# ipvsadm -ln
IP Virtual Server version 0.9.4 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 172.26.20.118:80 wlc
-> 172.26.20.114:80 Route 2 0 215767
TCP 172.26.20.118:10000 wlc
-> 172.26.20.114:10000 Route 1 0 0
TCP 172.26.20.118:21 wlc persistent 360
-> 172.26.20.114:21 Route 1 0 0
TCP 172.26.20.118:23 wlc
-> 172.26.20.114:23 Route 1 0 0
[root@koala ipvs]# ipvsadm -ln
IP Virtual Server version 0.9.4 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 172.26.20.118:80 wlc
-> 172.26.20.114:80 Route 2 0 221145
TCP 172.26.20.118:10000 wlc
-> 172.26.20.114:10000 Route 1 0 0
TCP 172.26.20.118:21 wlc persistent 360
-> 172.26.20.114:21 Route 1 0 0
TCP 172.26.20.118:23 wlc
-> 172.26.20.114:23 Route 1 0 0
[root@koala ipvs]# ipvsadm -ln
IP Virtual Server version 0.9.4 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 172.26.20.118:80 wlc
-> 172.26.20.114:80 Route 2 0 232129
TCP 172.26.20.118:10000 wlc
-> 172.26.20.114:10000 Route 1 0 0
TCP 172.26.20.118:21 wlc persistent 360
-> 172.26.20.114:21 Route 1 0 0
TCP 172.26.20.118:23 wlc
-> 172.26.20.114:23 Route 1 0 0
[root@koala ipvs]# vmstat
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 0 49768 1760 31736 0 0 12 5 3044 16 2 95 3
31 processes: 29 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 0.0% user, 96.4% system, 0.0% nice, 3.6% idle
Mem: 127268K av, 77636K used, 49632K free, 0K shrd, 1760K
buff
Swap: 136544K av, 0K used, 136544K free 31736K
cached
PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
3 root 20 0 0 0 0 SW 95.1 0.0 38:42 kapm-idled
The number of connections doesn't matter a lot, because the overhead of
hash table collision is not very high. I guess that there is something
wrong with your configuration/system, please check it.
Regards,
Wensong
On 11 Oct 2001, Hervé Guehl wrote:
>
> Hi, I made a little with lvs simply forward the http port with a lvs nat
> entry.. Then I use hammerhead (http://hammerhead.sourceforge.net/) to
> test the behaviour of the stuff under heavy load.
>
> My linux LVS box hanged (no kernel panic .. but the machine freezed, was
> around 28000 inactive connections).
> After reboot.. I could see in the log something that looks like a kind
> of "mempry overflow"..
>
> Here's what I could see :
>
> Oct 11 00:08:51 tds101 keepalived[1428]: TCP connection to
> [192.168.39.199:80] success.
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@(3) - end the application or thread (and invoke exit handlers)
> Tcl_DeleteExitHandler [Tcl_Finalize] (3) - end the application or
> thread (and invoke exit handlers)
> Tcl_DeleteExitHandler [Tcl_FinalizeThread] (3) - end the application or
> thread (and invoke exit handlers)
> Tcl_DeleteFileHandler (3) - associate procedure callbacks with files or
> devices (Unix only)
> Tcl_DeleteFileHandler [CrtFileHdlr] (3) - associate procedure callbacks
> with files or devices (Unix only)
> Tcl_DeleteFileHandler [Tcl_CreateFileHandler] (3) - associate procedure
> callbacks with files or devices (Unix only)
> Tcl_DeleteHashEntry (3) - procedures to manage hash tables
> Tcl_DeleteHashEntry [Hash] (3) - procedures to manage hash tables
>
> Tried with lvs 0.8.1 and lvs 0.9.4. Kernel is 2.4.10
>
> After recompiling ip_vs modules with "#define CONFIG_IP_VS_TAB_BITS
> 16" this behaviour was more difficult to obtain (I have only one
> "hammer" machine)..
>
> I dont know about the behaviour of LVS when reaching the max connection
> but would it be possible to drop new connections in this case and push
> something in syslog ? (as does iptables ..)
>
> Another thing seems, that packet unrelated to ipvs contrack are dropped
> without any message.. Would it be possible to have a message or to let
> the packets go thru the machine ??
>
> Regards
> Hervé
>
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users
>
|
