Hello All,

I'm wondering if someone might be able to point me in the direction of a problem with my HAProxy deployment. Under an ApacheBench load (40,000 concurrent) the backend stat on HAProxy on average shows around 22,000 connections whilst each of the three servers in the cluster is picking up about 6800 each. What is the backend statistic representing, delay in waiting for a server to open a connection? If so - what's a likely fix to improve the connection throughput on the servers?

Two of the servers are running LigHTTPd, and the other is using IIS.

Copy of the stats attached.

Cheers,

Robert.

Title: Statistics Report for HAProxy

HAProxy version 1.3.14.11, released 2008/12/04

Statistics Report for pid 19312


> General process information

pid = 19312 (process #1, nbproc = 1)
uptime = 0d 0h12m21s
system limits : memmax = unlimited ; ulimit-n = 1024576
maxsock = 80012
maxconn = 40000 (current conns = 1)

 active UP  backup UP
active UP, going down backup UP, going down
active DOWN, going up backup DOWN, going up
active or backup DOWN  not checked
Note: UP with load-balancing disabled is reported as "NOLB".
Display option:External ressources:
rps-test
QueueSessionsBytesDeniedErrorsWarningsServer
CurMaxLimitCurMaxLimitTotalLbTotInOutReqRespReqConnRespRetrRedisStatusWghtActBckChkDwnDwntmeThrtle
Frontend14000040000202167715966771364663253700847OPEN
Web100-09724-66083566083552208287196930630000 0no check35Y--
Web200-011130-74230174230158646681221209498000 0no check40Y--
Web300-011476-61766661765348797425228188744000 13no check41Y--
Backend0003233040000202078920207891596677136466325370000 13012m21s UP11630 00s


Reply via email to