Confirming mlockall is working?

Elastic Stack Elasticsearch
#1

Hey,
I've noticed some request time spikes with 0.16.2 that I hadn't
observed with 0.14.2. I know that mlockall was critical in ensuring
that GC didn't have much hit on response times.

The docs state that:
In order to see if this works or not, set the common.jna logging to
DEBUG level. A solution to “Unknown mlockall error 0” can be to set
ulimit -l unlimited.

I have set common.jna logging to DEBUG, but don't see any messages
logged from this class. Is that a confirmation that mlockall is
working or should I expect to see something indicating that it is
working?

Thanks!
Paul

#2

Yes, if you don't see anything in the lock, then it should be working.

On Wed, Jul 27, 2011 at 7:46 PM, ppearcy ppearcy@gmail.com wrote:

Hey,
I've noticed some request time spikes with 0.16.2 that I hadn't
observed with 0.14.2. I know that mlockall was critical in ensuring
that GC didn't have much hit on response times.

The docs state that:
In order to see if this works or not, set the common.jna logging to
DEBUG level. A solution to “Unknown mlockall error 0” can be to set
ulimit -l unlimited.

I have set common.jna logging to DEBUG, but don't see any messages
logged from this class. Is that a confirmation that mlockall is
working or should I expect to see something indicating that it is
working?

Thanks!
Paul

#3

Hi,

I had the same worries about whether my mlockall setting was actually
set, would it be possible to put a positive message in the logs like
"mlockall now working"?

Best Regards,

David.

On Jul 27, 6:01 pm, Shay Banon kim...@gmail.com wrote:

Yes, if you don't see anything in the lock, then it should be working.

On Wed, Jul 27, 2011 at 7:46 PM, ppearcy ppea...@gmail.com wrote:

Hey,
I've noticed some request time spikes with 0.16.2 that I hadn't
observed with 0.14.2. I know that mlockall was critical in ensuring
that GC didn't have much hit on response times.

The docs state that:
In order to see if this works or not, set the common.jna logging to
DEBUG level. A solution to “Unknown mlockall error 0” can be to set
ulimit -l unlimited.

I have set common.jna logging to DEBUG, but don't see any messages
logged from this class. Is that a confirmation that mlockall is
working or should I expect to see something indicating that it is
working?

Thanks!
Paul

#4

Yes, I think it makes sense. Something like, if its explicitly set, then log
a warn if its not working. Can you open an issue for it?

On Mon, Aug 1, 2011 at 2:56 PM, davrob2 daviroberts@gmail.com wrote:

Hi,

I had the same worries about whether my mlockall setting was actually
set, would it be possible to put a positive message in the logs like
"mlockall now working"?

Best Regards,

David.

On Jul 27, 6:01 pm, Shay Banon kim...@gmail.com wrote:

Yes, if you don't see anything in the lock, then it should be working.

On Wed, Jul 27, 2011 at 7:46 PM, ppearcy ppea...@gmail.com wrote:

Hey,
I've noticed some request time spikes with 0.16.2 that I hadn't
observed with 0.14.2. I know that mlockall was critical in ensuring
that GC didn't have much hit on response times.

The docs state that:
In order to see if this works or not, set the common.jna logging to
DEBUG level. A solution to “Unknown mlockall error 0” can be to set
ulimit -l unlimited.

I have set common.jna logging to DEBUG, but don't see any messages
logged from this class. Is that a confirmation that mlockall is
working or should I expect to see something indicating that it is
working?

Thanks!
Paul

#5

Issue raise here: https://github.com/elasticsearch/elasticsearch/issues/1194

thanks.

On Aug 1, 9:55 pm, Shay Banon kim...@gmail.com wrote:

Yes, I think it makes sense. Something like, if its explicitly set, then log
a warn if its not working. Can you open an issue for it?

On Mon, Aug 1, 2011 at 2:56 PM, davrob2 davirobe...@gmail.com wrote:

Hi,

I had the same worries about whether my mlockall setting was actually
set, would it be possible to put a positive message in the logs like
"mlockall now working"?

Best Regards,

David.

On Jul 27, 6:01 pm, Shay Banon kim...@gmail.com wrote:

Yes, if you don't see anything in the lock, then it should be working.

On Wed, Jul 27, 2011 at 7:46 PM, ppearcy ppea...@gmail.com wrote:

Hey,
I've noticed some request time spikes with 0.16.2 that I hadn't
observed with 0.14.2. I know that mlockall was critical in ensuring
that GC didn't have much hit on response times.

The docs state that:
In order to see if this works or not, set the common.jna logging to
DEBUG level. A solution to “Unknown mlockall error 0” can be to set
ulimit -l unlimited.

I have set common.jna logging to DEBUG, but don't see any messages
logged from this class. Is that a confirmation that mlockall is
working or should I expect to see something indicating that it is
working?

Thanks!
Paul

#6

I have a noob question here.

where (as in what file, is it etc/init.d/elasticsearch, or /etc/elasticsearch/elasticsearch.yml?) do I set the common.jna, and what is the syntax for setting it to debug?

do I just set
common.jna = DEBUG
next to
bootstrap.mlockall: true
?

Where would I then follow the log to see this? would that be in /var/log/elasticsearch/elasticsearch.log?

when would this appear there, when I start the service ("service elasticsearch start")?

#7