Re: Metrics in 6.5.1 names and stuff

2017-08-04 Thread Shalin Shekhar Mangar 
Hi Walter,

You need the metrics with the prefix "QUERY./srp" -- those will give
you request counts and percentiles for the /srp handler for a node.

Comments inline:

On Fri, Aug 4, 2017 at 3:23 AM, Walter Underwood  wrote:
> I’m trying to get what I want out of the metrics reporting in Solr. I want 
> the counts and percentiles for each request handler in each collection. If I 
> have “/srp”, “/suggest”, and “/seo”, I want three sets of metrics.
>
> I’m getting a lot of weird stuff. For counts for /srp in an eight node 
> cluster, one node (new-solr-c01) is reporting all of these graphite metric 
> names:
>
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c01.test3.cloud.cheggnet.com:8983/solr/questions_shard1_replica2/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c02.test3.cloud.cheggnet.com:8983/solr/questions_shard2_replica4/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c03.test3.cloud.cheggnet.com:8983/solr/questions_shard4_replica3/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c04.test3.cloud.cheggnet.com:8983/solr/questions_shard2_replica2/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c05.test3.cloud.cheggnet.com:8983/solr/questions_shard4_replica4/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c06.test3.cloud.cheggnet.com:8983/solr/questions_shard3_replica1/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c07.test3.cloud.cheggnet.com:8983/solr/questions_shard1_replica3/srp.post.requests.count
> new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c08.test3.cloud.cheggnet.com:8983/solr/questions_shard3_replica3/srp.post.requests.count
> new-solr-c01.solr.node.UPDATE.updateShardHandler./solr/questions/srp.get.requests.count
>
> 1. What are the metrics with the URLs? Are those distributed queries to other 
> shards? If so, why are they going to the local shard, too?

Yes, these are distributed queries to other shards. Any thing under
QUERY.httpShardHandler is either a distributed search request or an
admin request made by the overseer (in solrcloud).

The local shard isn't short circuited. A request to a local core (as
part of a distributed query) also goes through the HttpShardHandler.
This is why the local shard shows up here.

>
> 2. Why is the last one under UPDATE, when it is a query request handler?

It is not under just UPDATE. It is under UPDATE.updateShardHandler.
These are requests made using the UpdateShardHandler which is
typically used for distributed updates. I am not sure why a call was
made to the /srp handler using the update shard handler.

>
> 3. The jetty metrics seem to lump everything together under the HTTP method.

Is there a question here?

>
> 4. I enabled the http group, but don’t get any metrics from it.

The http group is not used currently. It was supposed to be used for
collecting metrics from http clients used inside Solr but since then
those have been folded into QUERY.httpShardHandler and
UPDATE.updateShardHandler metrics.

>
> Here is the config. The prefix is set to the nodename. For this test, I sent 
> the metrics to localhost and logged them with “nc -l 2003 > graphite.log”.
>
>
>   class="org.apache.solr.metrics.reporters.SolrGraphiteReporter">
>${graphite_host:NO_HOST_SET}
>2003
>${graphite_prefix:NO_PREFIX_SET}
>http://observer.wunderwood.org/  (my blog)
>
>



-- 
Regards,
Shalin Shekhar Mangar.

Metrics in 6.5.1 names and stuff

2017-08-03 Thread Walter Underwood 
I’m trying to get what I want out of the metrics reporting in Solr. I want the 
counts and percentiles for each request handler in each collection. If I have 
“/srp”, “/suggest”, and “/seo”, I want three sets of metrics.

I’m getting a lot of weird stuff. For counts for /srp in an eight node cluster, 
one node (new-solr-c01) is reporting all of these graphite metric names:

new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c01.test3.cloud.cheggnet.com:8983/solr/questions_shard1_replica2/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c02.test3.cloud.cheggnet.com:8983/solr/questions_shard2_replica4/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c03.test3.cloud.cheggnet.com:8983/solr/questions_shard4_replica3/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c04.test3.cloud.cheggnet.com:8983/solr/questions_shard2_replica2/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c05.test3.cloud.cheggnet.com:8983/solr/questions_shard4_replica4/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c06.test3.cloud.cheggnet.com:8983/solr/questions_shard3_replica1/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c07.test3.cloud.cheggnet.com:8983/solr/questions_shard1_replica3/srp.post.requests.count
new-solr-c01.solr.node.QUERY.httpShardHandler.http://new-solr-c08.test3.cloud.cheggnet.com:8983/solr/questions_shard3_replica3/srp.post.requests.count
new-solr-c01.solr.node.UPDATE.updateShardHandler./solr/questions/srp.get.requests.count

1. What are the metrics with the URLs? Are those distributed queries to other 
shards? If so, why are they going to the local shard, too?

2. Why is the last one under UPDATE, when it is a query request handler?

3. The jetty metrics seem to lump everything together under the HTTP method.

4. I enabled the http group, but don’t get any metrics from it.

Here is the config. The prefix is set to the nodename. For this test, I sent 
the metrics to localhost and logged them with “nc -l 2003 > graphite.log”.

   
 
   ${graphite_host:NO_HOST_SET}
   2003
   ${graphite_prefix:NO_PREFIX_SET}
   http://observer.wunderwood.org/  (my blog)