Mailing-List: contact user-help@cayenne.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cayenne.apache.org
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 9.2 \(3112\))
Subject: Re: Reconnecting automatically after DB downtime
From: Hugi Thordarson <hugi@karlmenn.is>
In-Reply-To: <135D44F4-0F99-4286-BF84-ED3404B1BF01@karlmenn.is>
Date: Mon, 7 Mar 2016 20:29:08 +0000
Content-Transfer-Encoding: quoted-printable
Message-Id: <E5A30D93-637E-490C-B54D-4FB2136EB360@karlmenn.is>
References: <15E01460-2B2C-4E0A-BE44-86D8020060F7@karlmenn.is>
 <DAF905B5-75AD-4B73-83FC-DCC1A4A66994@objectstyle.org>
 <586DD623-7C0F-4AF9-80C5-53DDB80A6ADA@karlmenn.is>
 <55049F21-3527-4B88-90BC-202D1451EDB7@objectstyle.org>
 <1CE7D336-E62F-443E-B6A1-4FC9863FC94F@karlmenn.is>
 <DC21E526-C64D-4438-97E3-0BF00C81CF03@objectstyle.org>
 <135D44F4-0F99-4286-BF84-ED3404B1BF01@karlmenn.is>
To: user@cayenne.apache.org

Is it possible it=E2=80=99s a configuration issue? I=E2=80=99m setting =
minConnections to 1 and maxConnections to 10. Should I rather leave =
these at default values or choose different values?

Cheers,
- hugi

// Hugi Thordarson
// http://www.loftfar.is/
// s. 895-6688


> On 7. mar. 2016, at 20:01, Hugi Thordarson <hugi@karlmenn.is> wrote:
>=20
> Thanks for the reply Andrus!
>=20
> Sounds like it might be the case. But actually, this now happens =
without the DB going down (and I don=E2=80=99t think it=E2=80=99s a =
communication problem with the DB either, since the DB runs on the same =
server as the apps).
>=20
> I recently added a PostCommitListener that does quite a lot of DB work =
and I believe the issue started around the same time. At first I thought =
it might be a deadlocking sitation, but a jstack thread dump from a dead =
process does not point to that.
>=20
> Is there any way for me to check the number and status of currently =
open connections?
>=20
> Cheers,
> - hugi
>=20
>=20
>=20
>> On 7. mar. 2016, at 18:54, Andrus Adamchik <andrus@objectstyle.org> =
wrote:
>>=20
>> Hmm... It shows there are zero connections in the pool. I have a =
theory about that. Looks like when you uncheck your first connection =
after the DB restart, pool goes through all connections to find a valid =
one, and removing bad connections in the process. As there are no longer =
valid connections there, the pool is drained completely. Then the error =
happens; and then at a later time, pool manager thread kicks in and =
fills the pool with new connections.=20
>>=20
>> I need to test it.. And perhaps move the validation to background.=20
>>=20
>> (and perhaps add Tomcat connection pool to a list of DataSource =
options ... that's the one I am using in Bootique: =
https://github.com/nhl/bootique-jdbc and elsewhere in production).
>>=20
>> Andrus
>>=20
>>> On Mar 7, 2016, at 2:46 AM, Hugi Thordarson <hugi@karlmenn.is> =
wrote:
>>>=20
>>> Hi all,
>>> This is still happening, even after I added a validationQuery, our =
app is dying quite frequently :(.  I=E2=80=99m not quite sure how to =
debug this, is there any way for me to catch where connections are being =
opened in the code and at what location they=E2=80=99re hanging?
>>>=20
>>> =
https://www.dropbox.com/s/8jkmh6513s6wwkn/Screenshot%202016-03-07%2010.21.=
53.png?dl=3D0
>>>=20
>>> Cheer,
>>> - hugi
>>>=20
>>> // Hugi Thordarson
>>> // http://www.loftfar.is/
>>> // s. 895-6688
>>>=20
>>>=20
>>>=20
>>>> On 29. feb. 2016, at 11:25, Andrus Adamchik =
<andrus@objectstyle.org> wrote:
>>>>=20
>>>>=20
>>>>> On Feb 29, 2016, at 2:20 PM, Hugi Thordarson <hugi@karlmenn.is> =
wrote:
>>>>>=20
>>>>> What does validationQuery do?
>>>>=20
>>>> Periodically executes for each pooled connection, and kills =
connections that throw during validation. So it ensures that all pooled =
connections are in a good state.
>>>>=20
>>>> Andrus
>>>>=20
>>>=20
>>=20
>=20