loadbalancer errors, can you please explain?

Mikra

Active Member
#1
Hello,

Can anyone please explain when exactly these errors occur?

1. ExtConn timed out while connecting.
2. Failure detected: Connection Failure, 110:Connection timed out
3. PingConn timed out while connecting.
4. External processor is not available!
5. Fail all outstanding requests!

We're trying to debug the fact that google webmaster tools is reporting an increase in 503 errors since we installed LSLB in front of an already existing webserver.

LSLB is reporting those errors regularly in the logs but I'm not sure why that is. At first I thought of a network issue but it turns out not to be the case. I had a ping running from the LSLB box to the webserver box and had 0% packet loss while LSLB reported errors.

Maybe someone can explain to me when the above errors happen so we can, hopefully, identify the problem.

I'm thinking that maybe LSLB thinks the webserver is dead at some point (maybe because of a slow loading page, timing out?) and the "fail all outstanding requests" error might cause google to receive 503 errors too?

I'm not sure what the errors mean exactly and when they happen. Hopefully someone here can help us out.
 

Mikra

Active Member
#3
I have not whitelisted anything. Where exactly can I do that?
My per client throttling is like this on the webserver:

Static Requests/second 0
Dynamic Requests/second 0
Outbound Bandwidth (bytes/sec) 0
Inbound Bandwidth (bytes/sec) 0
Connection Soft Limit 10000
Connection Hard Limit 10000
Grace Period (sec) 15
Banned Period (sec) 300
 

NiteWave

Administrator
#4
this is the default setting and almost unlimited connections -- so white list looks unnecessary.

is there any valuable error message at web server around the time stamp when lslb have the connection errors?
 

Mikra

Active Member
#6
It's a bug in litespeed load balancer, it seems.

We have switched to 'haproxy' and all the problems are gone now. And I must say, haproxy is a MAYOR improvement too. We're not going to switch back to lslb.

LSLB was running on a VPS and was constantly running at about 70% CPU load during the day (load balancer was also doing gzip compressing, offloading the webservers). VMWare stats show 70% CPU load every single day in the last 8 months.
Now the same VPS is running haproxy and the CPU load dropped from 70% to 20%. We no longer have those 503 errors and haproxy is even using less memory.

I think you guy should consider comparing your product to haproxy now. haproxy is opensource, has more features, easy configurable and performs better. So, right now, there is no reason for us to switch back to lslb.
 
Top