|
If you were logged in you would be able to see more operations.
|
|
|
| Resolution Date: |
16/Aug/07 07:14 PM
|
|
In the ServerContext factory the same thread pool is being used to process I/O
and request handling. This is depleting the thread pool rapidly causing the
server to hang under certain conditions. I caught this bug on a x86_64 machine
with multiple cores when the ReferralITest test was hanging. The right number
of connections and requests being processed could have caused this issue to
arrise. Regardless thankfully the issue appeared and Trustin found the cause
for it.
Trustin tracked the bug down to the way we misconfigured MINA's thread pools.
He is currently fixing the problem in trunk and I suspect the same issue exists
within the 1.0 branch.
I suspect this could have caused serious issues in performance if not causing
several load test clients to hang while others had requests processed. I
recommend reassessing performance metrics once again after this fix. There
should be a considerable boost. I think our impression that MINA was a
bottleneck may have been attributed to this misconfiguration issue.
Anyways once Trustin has the fix in trunk I will apply it to the 1.0 branch and
make sure all tests pass successfully. I am assigning this issue to myself
even though T is providing the fix so I can manage the application on the
1.0 branch and document the results.
|
|
Description
|
In the ServerContext factory the same thread pool is being used to process I/O
and request handling. This is depleting the thread pool rapidly causing the
server to hang under certain conditions. I caught this bug on a x86_64 machine
with multiple cores when the ReferralITest test was hanging. The right number
of connections and requests being processed could have caused this issue to
arrise. Regardless thankfully the issue appeared and Trustin found the cause
for it.
Trustin tracked the bug down to the way we misconfigured MINA's thread pools.
He is currently fixing the problem in trunk and I suspect the same issue exists
within the 1.0 branch.
I suspect this could have caused serious issues in performance if not causing
several load test clients to hang while others had requests processed. I
recommend reassessing performance metrics once again after this fix. There
should be a considerable boost. I think our impression that MINA was a
bottleneck may have been attributed to this misconfiguration issue.
Anyways once Trustin has the fix in trunk I will apply it to the 1.0 branch and
make sure all tests pass successfully. I am assigning this issue to myself
even though T is providing the fix so I can manage the application on the
1.0 branch and document the results.
|
Show » |
|
Bug
Closed
Blocker
Using same thread pool for I/O processing and request handling causes clients to hang and MINA thread pool depletion
