You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by Jeremy Levy <je...@gmail.com> on 2012/08/28 16:44:59 UTC
tmp-storage cleanup with fileQueueCursor
When using fileQueueCursor, as I understand it the data is written to the
tmp-storage directory. When does this directory get cleaned up. For
example I tested putting about 3 million messages on a queue and it
successfully wrote them to disk (as opposed to memory). It took up about
8.7GB. I then deleted the queue from the web console, however the 8.7GB
remained. Is there a job or something that I need to configure for this
space to be reclaimed?
Thanks.
Jeremy
<destinationPolicy>
<policyMap>
<policyEntries>
<policyEntry topic=">" producerFlowControl="false"
memoryLimit="2mb">
<pendingSubscriberPolicy>
<vmCursor />
</pendingSubscriberPolicy>
</policyEntry>
<policyEntry queue=">" producerFlowControl="false"
memoryLimit="2mb" prioritizedMessages="true">
<!-- Use VM cursor for better latency
For more information, see:
http://activemq.apache.org/message-cursors.html
-->
<pendingQueuePolicy>
<!--vmQueueCursor/-->
<fileQueueCursor/>
</pendingQueuePolicy>
</policyEntry>
</policyEntries>
</policyMap>
</destinationPolicy>
--
Jeremy Levy
Re: tmp-storage cleanup with fileQueueCursor
Posted by Gary Tully <ga...@gmail.com>.
thanks for reporting back and closing the loop on this one :-)
On 28 August 2012 22:55, Jeremy Levy <je...@gmail.com> wrote:
> I was able to solve both issues, It wasn't clear to me that
> memoryLimit="2mb" was the corresponding setting on the queue policy in
> regards to the memory limit threshold before spooling to disk. Therefore I
> think it was starting with the disk right away. After changing that to
> something reasonably higher it now works as expected for both cleaning up
> the disk and spooling.
>
> Hope someone else finds this useful and saves them the trouble.
>
> Jeremy
>
>
>
> On Tue, Aug 28, 2012 at 12:08 PM, Jeremy Levy <je...@gmail.com> wrote:
>
>> Also, it doesn't seem as though fileQueueCursor is working as described in
>> many places in the documentation:
>>
>> http://activemq.apache.org/message-cursors.html says:
>> "When memory in the broker reaches its limit, it can page messages to
>> temporary files on disk. "
>>
>> http://activemq.apache.org/javalangoutofmemory.html says:
>> "The Broker (using Message Cursors<http://activemq.apache.org/message-cursors.html>)
>> will spool non-persistent messages to disk when the default memory usage
>> threshold for a destination is reached. This threshold value is specified
>> to the Broker via the <memoryUsage> section of the <systemUsage>
>> configuration in activemq.xml<http://activemq.apache.org/xml-configuration.html>.
>> This feature allows producers to continue sending messages when there are
>> slow consumers without exhausting available memory or reverting to producer
>> flow control <http://activemq.apache.org/producer-flow-control.html>."
>>
>> I've tested this with one queue, configured in with <fileQueueCursor/> and
>> various <memoryUsage> configurations. Messages appear to always be
>> written to disk. This, combined with the tmp-storage not being clean up
>> means that my activemq server is a ticking time bomb until it runs out of
>> disk space.
>>
>> I could get away with the disk space not being cleaned out if it was only
>> writing to disk once the memory limit was reached but that doesn't seem to
>> be the case. Any suggestions?
>>
>> Jeremy
>>
>> On Tue, Aug 28, 2012 at 11:14 AM, Jeremy Levy <je...@gmail.com> wrote:
>>
>>> Hi Gary-
>>>
>>> Thanks for the quick response, as I read that it's fixed in 5.6.0,
>>> however I'm using 5.6.0... I'm only seeing the directory cleaned up during
>>> a restart.
>>>
>>> Jeremy
>>>
>>> On Tue, Aug 28, 2012 at 11:03 AM, Gary Tully <ga...@gmail.com>wrote:
>>>
>>>> If you are sending non persistent messages see
>>>> https://issues.apache.org/jira/browse/AMQ-3780
>>>>
>>>> On 28 August 2012 15:44, Jeremy Levy <je...@gmail.com> wrote:
>>>> > When using fileQueueCursor, as I understand it the data is written to
>>>> the
>>>> > tmp-storage directory. When does this directory get cleaned up. For
>>>> > example I tested putting about 3 million messages on a queue and it
>>>> > successfully wrote them to disk (as opposed to memory). It took up
>>>> about
>>>> > 8.7GB. I then deleted the queue from the web console, however the
>>>> 8.7GB
>>>> > remained. Is there a job or something that I need to configure for
>>>> this
>>>> > space to be reclaimed?
>>>> >
>>>> > Thanks.
>>>> >
>>>> > Jeremy
>>>> >
>>>> >
>>>> > <destinationPolicy>
>>>> > <policyMap>
>>>> > <policyEntries>
>>>> > <policyEntry topic=">" producerFlowControl="false"
>>>> > memoryLimit="2mb">
>>>> > <pendingSubscriberPolicy>
>>>> > <vmCursor />
>>>> > </pendingSubscriberPolicy>
>>>> > </policyEntry>
>>>> >
>>>> > <policyEntry queue=">" producerFlowControl="false"
>>>> > memoryLimit="2mb" prioritizedMessages="true">
>>>> > <!-- Use VM cursor for better latency
>>>> > For more information, see:
>>>> >
>>>> > http://activemq.apache.org/message-cursors.html
>>>> > -->
>>>> > <pendingQueuePolicy>
>>>> > <!--vmQueueCursor/-->
>>>> > <fileQueueCursor/>
>>>> > </pendingQueuePolicy>
>>>> > </policyEntry>
>>>> >
>>>> > </policyEntries>
>>>> > </policyMap>
>>>> > </destinationPolicy>
>>>> >
>>>> >
>>>> > --
>>>> > Jeremy Levy
>>>>
>>>>
>>>>
>>>> --
>>>> http://fusesource.com
>>>> http://blog.garytully.com
>>>>
>>>
>>>
>>>
>>> --
>>> Jeremy Levy
>>>
>>
>>
>>
>> --
>> Jeremy Levy
>>
>
>
>
> --
> Jeremy Levy
--
http://fusesource.com
http://blog.garytully.com
Re: tmp-storage cleanup with fileQueueCursor
Posted by Jeremy Levy <je...@gmail.com>.
I was able to solve both issues, It wasn't clear to me that
memoryLimit="2mb" was the corresponding setting on the queue policy in
regards to the memory limit threshold before spooling to disk. Therefore I
think it was starting with the disk right away. After changing that to
something reasonably higher it now works as expected for both cleaning up
the disk and spooling.
Hope someone else finds this useful and saves them the trouble.
Jeremy
On Tue, Aug 28, 2012 at 12:08 PM, Jeremy Levy <je...@gmail.com> wrote:
> Also, it doesn't seem as though fileQueueCursor is working as described in
> many places in the documentation:
>
> http://activemq.apache.org/message-cursors.html says:
> "When memory in the broker reaches its limit, it can page messages to
> temporary files on disk. "
>
> http://activemq.apache.org/javalangoutofmemory.html says:
> "The Broker (using Message Cursors<http://activemq.apache.org/message-cursors.html>)
> will spool non-persistent messages to disk when the default memory usage
> threshold for a destination is reached. This threshold value is specified
> to the Broker via the <memoryUsage> section of the <systemUsage>
> configuration in activemq.xml<http://activemq.apache.org/xml-configuration.html>.
> This feature allows producers to continue sending messages when there are
> slow consumers without exhausting available memory or reverting to producer
> flow control <http://activemq.apache.org/producer-flow-control.html>."
>
> I've tested this with one queue, configured in with <fileQueueCursor/> and
> various <memoryUsage> configurations. Messages appear to always be
> written to disk. This, combined with the tmp-storage not being clean up
> means that my activemq server is a ticking time bomb until it runs out of
> disk space.
>
> I could get away with the disk space not being cleaned out if it was only
> writing to disk once the memory limit was reached but that doesn't seem to
> be the case. Any suggestions?
>
> Jeremy
>
> On Tue, Aug 28, 2012 at 11:14 AM, Jeremy Levy <je...@gmail.com> wrote:
>
>> Hi Gary-
>>
>> Thanks for the quick response, as I read that it's fixed in 5.6.0,
>> however I'm using 5.6.0... I'm only seeing the directory cleaned up during
>> a restart.
>>
>> Jeremy
>>
>> On Tue, Aug 28, 2012 at 11:03 AM, Gary Tully <ga...@gmail.com>wrote:
>>
>>> If you are sending non persistent messages see
>>> https://issues.apache.org/jira/browse/AMQ-3780
>>>
>>> On 28 August 2012 15:44, Jeremy Levy <je...@gmail.com> wrote:
>>> > When using fileQueueCursor, as I understand it the data is written to
>>> the
>>> > tmp-storage directory. When does this directory get cleaned up. For
>>> > example I tested putting about 3 million messages on a queue and it
>>> > successfully wrote them to disk (as opposed to memory). It took up
>>> about
>>> > 8.7GB. I then deleted the queue from the web console, however the
>>> 8.7GB
>>> > remained. Is there a job or something that I need to configure for
>>> this
>>> > space to be reclaimed?
>>> >
>>> > Thanks.
>>> >
>>> > Jeremy
>>> >
>>> >
>>> > <destinationPolicy>
>>> > <policyMap>
>>> > <policyEntries>
>>> > <policyEntry topic=">" producerFlowControl="false"
>>> > memoryLimit="2mb">
>>> > <pendingSubscriberPolicy>
>>> > <vmCursor />
>>> > </pendingSubscriberPolicy>
>>> > </policyEntry>
>>> >
>>> > <policyEntry queue=">" producerFlowControl="false"
>>> > memoryLimit="2mb" prioritizedMessages="true">
>>> > <!-- Use VM cursor for better latency
>>> > For more information, see:
>>> >
>>> > http://activemq.apache.org/message-cursors.html
>>> > -->
>>> > <pendingQueuePolicy>
>>> > <!--vmQueueCursor/-->
>>> > <fileQueueCursor/>
>>> > </pendingQueuePolicy>
>>> > </policyEntry>
>>> >
>>> > </policyEntries>
>>> > </policyMap>
>>> > </destinationPolicy>
>>> >
>>> >
>>> > --
>>> > Jeremy Levy
>>>
>>>
>>>
>>> --
>>> http://fusesource.com
>>> http://blog.garytully.com
>>>
>>
>>
>>
>> --
>> Jeremy Levy
>>
>
>
>
> --
> Jeremy Levy
>
--
Jeremy Levy
Re: tmp-storage cleanup with fileQueueCursor
Posted by Jeremy Levy <je...@gmail.com>.
Also, it doesn't seem as though fileQueueCursor is working as described in
many places in the documentation:
http://activemq.apache.org/message-cursors.html says:
"When memory in the broker reaches its limit, it can page messages to
temporary files on disk. "
http://activemq.apache.org/javalangoutofmemory.html says:
"The Broker (using Message
Cursors<http://activemq.apache.org/message-cursors.html>)
will spool non-persistent messages to disk when the default memory usage
threshold for a destination is reached. This threshold value is specified
to the Broker via the <memoryUsage> section of the <systemUsage>
configuration in
activemq.xml<http://activemq.apache.org/xml-configuration.html>.
This feature allows producers to continue sending messages when there are
slow consumers without exhausting available memory or reverting to producer
flow control <http://activemq.apache.org/producer-flow-control.html>."
I've tested this with one queue, configured in with <fileQueueCursor/> and
various <memoryUsage> configurations. Messages appear to always be written
to disk. This, combined with the tmp-storage not being clean up means that
my activemq server is a ticking time bomb until it runs out of disk space.
I could get away with the disk space not being cleaned out if it was only
writing to disk once the memory limit was reached but that doesn't seem to
be the case. Any suggestions?
Jeremy
On Tue, Aug 28, 2012 at 11:14 AM, Jeremy Levy <je...@gmail.com> wrote:
> Hi Gary-
>
> Thanks for the quick response, as I read that it's fixed in 5.6.0, however
> I'm using 5.6.0... I'm only seeing the directory cleaned up during a
> restart.
>
> Jeremy
>
> On Tue, Aug 28, 2012 at 11:03 AM, Gary Tully <ga...@gmail.com> wrote:
>
>> If you are sending non persistent messages see
>> https://issues.apache.org/jira/browse/AMQ-3780
>>
>> On 28 August 2012 15:44, Jeremy Levy <je...@gmail.com> wrote:
>> > When using fileQueueCursor, as I understand it the data is written to
>> the
>> > tmp-storage directory. When does this directory get cleaned up. For
>> > example I tested putting about 3 million messages on a queue and it
>> > successfully wrote them to disk (as opposed to memory). It took up
>> about
>> > 8.7GB. I then deleted the queue from the web console, however the 8.7GB
>> > remained. Is there a job or something that I need to configure for this
>> > space to be reclaimed?
>> >
>> > Thanks.
>> >
>> > Jeremy
>> >
>> >
>> > <destinationPolicy>
>> > <policyMap>
>> > <policyEntries>
>> > <policyEntry topic=">" producerFlowControl="false"
>> > memoryLimit="2mb">
>> > <pendingSubscriberPolicy>
>> > <vmCursor />
>> > </pendingSubscriberPolicy>
>> > </policyEntry>
>> >
>> > <policyEntry queue=">" producerFlowControl="false"
>> > memoryLimit="2mb" prioritizedMessages="true">
>> > <!-- Use VM cursor for better latency
>> > For more information, see:
>> >
>> > http://activemq.apache.org/message-cursors.html
>> > -->
>> > <pendingQueuePolicy>
>> > <!--vmQueueCursor/-->
>> > <fileQueueCursor/>
>> > </pendingQueuePolicy>
>> > </policyEntry>
>> >
>> > </policyEntries>
>> > </policyMap>
>> > </destinationPolicy>
>> >
>> >
>> > --
>> > Jeremy Levy
>>
>>
>>
>> --
>> http://fusesource.com
>> http://blog.garytully.com
>>
>
>
>
> --
> Jeremy Levy
>
--
Jeremy Levy
Re: tmp-storage cleanup with fileQueueCursor
Posted by Jeremy Levy <je...@gmail.com>.
Hi Gary-
Thanks for the quick response, as I read that it's fixed in 5.6.0, however
I'm using 5.6.0... I'm only seeing the directory cleaned up during a
restart.
Jeremy
On Tue, Aug 28, 2012 at 11:03 AM, Gary Tully <ga...@gmail.com> wrote:
> If you are sending non persistent messages see
> https://issues.apache.org/jira/browse/AMQ-3780
>
> On 28 August 2012 15:44, Jeremy Levy <je...@gmail.com> wrote:
> > When using fileQueueCursor, as I understand it the data is written to the
> > tmp-storage directory. When does this directory get cleaned up. For
> > example I tested putting about 3 million messages on a queue and it
> > successfully wrote them to disk (as opposed to memory). It took up about
> > 8.7GB. I then deleted the queue from the web console, however the 8.7GB
> > remained. Is there a job or something that I need to configure for this
> > space to be reclaimed?
> >
> > Thanks.
> >
> > Jeremy
> >
> >
> > <destinationPolicy>
> > <policyMap>
> > <policyEntries>
> > <policyEntry topic=">" producerFlowControl="false"
> > memoryLimit="2mb">
> > <pendingSubscriberPolicy>
> > <vmCursor />
> > </pendingSubscriberPolicy>
> > </policyEntry>
> >
> > <policyEntry queue=">" producerFlowControl="false"
> > memoryLimit="2mb" prioritizedMessages="true">
> > <!-- Use VM cursor for better latency
> > For more information, see:
> >
> > http://activemq.apache.org/message-cursors.html
> > -->
> > <pendingQueuePolicy>
> > <!--vmQueueCursor/-->
> > <fileQueueCursor/>
> > </pendingQueuePolicy>
> > </policyEntry>
> >
> > </policyEntries>
> > </policyMap>
> > </destinationPolicy>
> >
> >
> > --
> > Jeremy Levy
>
>
>
> --
> http://fusesource.com
> http://blog.garytully.com
>
--
Jeremy Levy
Re: tmp-storage cleanup with fileQueueCursor
Posted by Gary Tully <ga...@gmail.com>.
If you are sending non persistent messages see
https://issues.apache.org/jira/browse/AMQ-3780
On 28 August 2012 15:44, Jeremy Levy <je...@gmail.com> wrote:
> When using fileQueueCursor, as I understand it the data is written to the
> tmp-storage directory. When does this directory get cleaned up. For
> example I tested putting about 3 million messages on a queue and it
> successfully wrote them to disk (as opposed to memory). It took up about
> 8.7GB. I then deleted the queue from the web console, however the 8.7GB
> remained. Is there a job or something that I need to configure for this
> space to be reclaimed?
>
> Thanks.
>
> Jeremy
>
>
> <destinationPolicy>
> <policyMap>
> <policyEntries>
> <policyEntry topic=">" producerFlowControl="false"
> memoryLimit="2mb">
> <pendingSubscriberPolicy>
> <vmCursor />
> </pendingSubscriberPolicy>
> </policyEntry>
>
> <policyEntry queue=">" producerFlowControl="false"
> memoryLimit="2mb" prioritizedMessages="true">
> <!-- Use VM cursor for better latency
> For more information, see:
>
> http://activemq.apache.org/message-cursors.html
> -->
> <pendingQueuePolicy>
> <!--vmQueueCursor/-->
> <fileQueueCursor/>
> </pendingQueuePolicy>
> </policyEntry>
>
> </policyEntries>
> </policyMap>
> </destinationPolicy>
>
>
> --
> Jeremy Levy
--
http://fusesource.com
http://blog.garytully.com