cmullaparthi
commented
10 years ago Thanks for the bug report.
I've just finished redesigning the pipelining as another user of ibrowse had similar issues. This works much better. I'll send you some patches later today. Would appreciate it if you can try and tell me if it works for you.
W: http://chandrusoft.wordpress.com
On 24 Jul 2014, at 09:40, coolfeature notifications@github.com wrote:
I have a webservice that for every HTTP post does three ibrowse HTTP post requests to other webservices, waits for them to return, processes responses and then returns back.
Everything seems to work fine at first, but then after a day or so, especially when the service has to deal with a burst in requests I start getting many ibrowse timeouts.
We are not talking about huge volumes here. The service processes on average 50 req/s. The peaks can go up to 200 req/s. Whenever this happens the timeout issue starts showing up.
Here is some load balancing information when everything works fine:
Server:port | ETS | Num conns | LB Pid
abc:443 | 93394 | 45 | <0.795.0> def3:443 | 90936 | 10 | <0.787.0> ghi:443 | 92575 | 20 | <0.792.0>CONNECTIONS EST: 87 CONNECTIONS TIME_WAIT: 316
And this is when we are experiencing the issue:
Server:port | ETS | Num conns | LB Pid
abc:443 | 15348 | 182 | <0.23014.250> def:443 | 90936 | 193 | <0.787.0> ghi:443 | 16143 | 129 | <0.19076.250>CONNECTIONS ESTABLISHED: 551 CONNECTIONS WAITING: 25
The connections waiting shows that when there is an issue, the service goes to the EST state and then hangs on it too long therefore generating a timeout.
I am using 5000 for both max_session and pipline_size. Perhaps tweaking those values may help. The only when to get back on track is to restart the webservice.
Is anyone aware of any known limitations of ibrowse and timeout related issues?
— Reply to this email directly or view it on GitHub.
spc16670
I have a webservice that for every HTTP post does three ibrowse HTTP post requests to other webservices, waits for them to return, processes responses and then returns back.
Everything seems to work fine at first, but then after a day or so, especially when the service has to deal with a burst in requests I start getting many ibrowse timeouts.
We are not talking about huge volumes here. The service processes on average 50 req/s. The peaks can go up to 200 req/s. Whenever this happens the timeout issue starts showing up.
Here is some load balancing information when everything works fine:
Server:port | ETS | Num conns | LB Pid
CONNECTIONS EST: 87 CONNECTIONS TIME_WAIT: 316
And this is when we are experiencing the issue:
Server:port | ETS | Num conns | LB Pid
CONNECTIONS ESTABLISHED: 551 CONNECTIONS WAITING: 25
The connections waiting shows that when there is an issue, the service goes to the EST state and then hangs on it too long therefore generating a timeout.
I am using 5000 for both max_session and pipline_size. Perhaps tweaking those values may help. The only way to get back on track is to restart the webservice.
Is anyone aware of any known limitations of ibrowse and timeout related issues?