| erik |
Thu Nov 15 16:02:48 GMT 2007 |
| |
Hello again,
I have just released a patch on one of our web-applications with primrose as a new pool-controller. This web-application is under heavy load from thousands of users at the same time. We have one pool per web-server. I had the pool configured to 10 connections base with 10 overflow per server. The log showed that the pool was filled with 10 connections, but reduced this number to around 3 on average since no more was needed. This is a some of the regular lines the pool-logs printed:
[Pool-MPSoap] [available(1), active(1), overflow(10)]
[Pool-MPSoap] [available(2), active(1), overflow(10)]
[Pool-MPSoap] [available(1), active(2), overflow(10)]
etc...
Then we did some restarts of our servers and the pool was again filled with 10 connections, and slowly decreased the number to around 2-3 as before.
Then suddenly some strange loglines appeared.
[Pool-MPSoap] [available(2), active(0), overflow(10)]
extraction connection blablabl |
| |
| erik |
Thu Nov 15 16:06:50 GMT 2007 |
| |
| oh by the way. The problem accured after enabling the checkSQL in the admin properties file. |
| sedj |
Thu Nov 15 18:44:37 GMT 2007 |
| |
Well its hard to say from a few lines of pasted log, but I would guess that there really are that number of connections being used.
Its possible that enabling the checkSQL is slowing down the returning of connections, because it is processing another statement.
I'm not convinced its a bug per-se, because primrose 2 has been out there for many years now, and a bug that was causing such a heavy use of connections would be reported by more than one user.
If you are new to primrose, I would suggest looking at Primrose 3 - its a cleaner implementation, and you may get better results (note that current version is 3.0.7, a new release, 3.0.8 will be out in a day or two which will handle you stopping and starting your database quite gracefully). |
| erik |
Fri Nov 16 10:24:29 GMT 2007 |
| |
Thanks for your reply.
Tonight we encountered some major problems and had to make a rollback to our old connection-pool. I'm not yet sure what the problem really was, but our clients using our SOAP-services encoutered time-outs on a lot of requests.
Yes i am new to primrose, and it is possible i am not using it correct. If you have the time and/or interest i would really like you to take a look at the logs produced by primrose. I can email them to you or I can try to paste some lines here, but then I will have to slice them down a bit.
What is worrying me is this:
For this particular service we are running two seperate webservers with a loadbalancer between them, each server with its own connectionpool. I have checked our request-logs and the loadbalancer worked fine with aproximatly 50-50 split of requests between them.
Server 1 and its pool was all fine. With these loglines printed by primrose throughout the day and night:
[Pool-MPSoap] [available(1), |
| sedj |
Sat Nov 17 10:16:28 GMT 2007 |
| |
Yeah, that is a bit odd. Have never seen anything like it.
Can you mail me the errant pool log, my address is "sedj" at this domain. |
|
