You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by sebb <se...@gmail.com> on 2010/01/15 17:04:27 UTC
[DBCP] Continuum failure - increase wait time?
It's looking as if the problem with the Continuum test failures is
probably because the system is overloaded.
Perhaps it does not make sense to have such a short wait time, i.e. 100ms.
Maybe better to increase the time to (say) 500ms, and keep the hold
time at 2 * wait.
I'm going to try this to see.
It will obviously increase the time it takes to run the test, but it
should not take much more than 1s or 1.5s
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [DBCP] Continuum failure - increase wait time?
Posted by sebb <se...@gmail.com>.
On 16/01/2010, Phil Steitz <ph...@gmail.com> wrote:
> sebb wrote:
> > On 15/01/2010, sebb <se...@gmail.com> wrote:
> >> It's looking as if the problem with the Continuum test failures is
> >> probably because the system is overloaded.
> >>
> >> Perhaps it does not make sense to have such a short wait time, i.e. 100ms.
> >>
> >> Maybe better to increase the time to (say) 500ms, and keep the hold
> >> time at 2 * wait.
> >>
> >> I'm going to try this to see.
> >> It will obviously increase the time it takes to run the test, but it
> >> should not take much more than 1s or 1.5s
> >
> > I'm fairly confident that the only test failures now are due to slow
> > running on the overloaded Continuum host. The last 15 DBCP runs have
> > been successful. So might be good to consider creating a new release
> > candidate.
> >
> > However, before that, can we perhaps fix:
> >
> > DBCP-231 SharedPoolDataSource does not need userKeys LRUMap cache
> >
> > This would solve the question as to whether the correct entry is being
> > removed currently, and would obviously avoid any possibility of stale
> > entries.
>
>
> Agreed. I think we can safely eliminate the "cache" as I suspect
> with anything but the most ancient JDKs, the overhead associated
> with the cache itself swamps any benefit it provides. We can also
> eliminate the Collections LRUMap class by doing this.
Except if you want to use it for the PerUserPoolDataSource cache ;-)
> Phil
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> > For additional commands, e-mail: dev-help@commons.apache.org
> >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [DBCP] Continuum failure - increase wait time?
Posted by Phil Steitz <ph...@gmail.com>.
sebb wrote:
> On 15/01/2010, sebb <se...@gmail.com> wrote:
>> It's looking as if the problem with the Continuum test failures is
>> probably because the system is overloaded.
>>
>> Perhaps it does not make sense to have such a short wait time, i.e. 100ms.
>>
>> Maybe better to increase the time to (say) 500ms, and keep the hold
>> time at 2 * wait.
>>
>> I'm going to try this to see.
>> It will obviously increase the time it takes to run the test, but it
>> should not take much more than 1s or 1.5s
>
> I'm fairly confident that the only test failures now are due to slow
> running on the overloaded Continuum host. The last 15 DBCP runs have
> been successful. So might be good to consider creating a new release
> candidate.
>
> However, before that, can we perhaps fix:
>
> DBCP-231 SharedPoolDataSource does not need userKeys LRUMap cache
>
> This would solve the question as to whether the correct entry is being
> removed currently, and would obviously avoid any possibility of stale
> entries.
Agreed. I think we can safely eliminate the "cache" as I suspect
with anything but the most ancient JDKs, the overhead associated
with the cache itself swamps any benefit it provides. We can also
eliminate the Collections LRUMap class by doing this.
Phil
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org
Re: [DBCP] Continuum failure - increase wait time?
Posted by sebb <se...@gmail.com>.
On 15/01/2010, sebb <se...@gmail.com> wrote:
> It's looking as if the problem with the Continuum test failures is
> probably because the system is overloaded.
>
> Perhaps it does not make sense to have such a short wait time, i.e. 100ms.
>
> Maybe better to increase the time to (say) 500ms, and keep the hold
> time at 2 * wait.
>
> I'm going to try this to see.
> It will obviously increase the time it takes to run the test, but it
> should not take much more than 1s or 1.5s
I'm fairly confident that the only test failures now are due to slow
running on the overloaded Continuum host. The last 15 DBCP runs have
been successful. So might be good to consider creating a new release
candidate.
However, before that, can we perhaps fix:
DBCP-231 SharedPoolDataSource does not need userKeys LRUMap cache
This would solve the question as to whether the correct entry is being
removed currently, and would obviously avoid any possibility of stale
entries.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org