Intermittently search slow response


(Jason Wee) #1

Hi, we are experience a strange issue, it happen intermittently. we have
three elasticsearch nodes running version 0.90.7 and started a new cluster.
Then as data coming in, we query one of the nodes, sometime the result
return immediately. Sometime it just hang there forever until ctrl+z is
issued. Not exactly sure what is wrong. Can anyone pinpoint where might be
the possible cause?

[user@node1 ~]$ time curl -XGET
'http://localhost:9200/index_test/_search?q=key:keywords&pretty=true'
{
"took" : 69,
"timed_out" : false,
"_shards" : {
"total" : 4500,
"successful" : 4500,
"failed" : 0
},
"hits" : {
"total" : 2201,
"max_score" : 6.192957,
"hits" : [ {
"_index" : "index_test",
...
...
...
real 0m0.072s
user 0m0.001s
sys 0m0.003s

[user@node1 ~]$ time curl -XGET
'http://localhost:9200/index_test/_search?q=key:keywords&pretty=true'

^C

real 16m34.575s
user 0m0.006s
sys 0m0.012s

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/868a0a1f-ee19-4a2d-807c-1379f023e9f2%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(simonw-2) #2

you are searching on 4500 shards? Is that right? I am just asking there
could be a lot of stuff going on though.

simon
On Friday, November 29, 2013 11:51:35 AM UTC+1, Jason Wee wrote:

Hi, we are experience a strange issue, it happen intermittently. we have
three elasticsearch nodes running version 0.90.7 and started a new cluster.
Then as data coming in, we query one of the nodes, sometime the result
return immediately. Sometime it just hang there forever until ctrl+z is
issued. Not exactly sure what is wrong. Can anyone pinpoint where might be
the possible cause?

[user@node1 ~]$ time curl -XGET '
http://localhost:9200/index_test/_search?q=key:keywords&pretty=true'
{
"took" : 69,
"timed_out" : false,
"_shards" : {
"total" : 4500,
"successful" : 4500,
"failed" : 0
},
"hits" : {
"total" : 2201,
"max_score" : 6.192957,
"hits" : [ {
"_index" : "index_test",
...
...
...
real 0m0.072s
user 0m0.001s
sys 0m0.003s

[user@node1 ~]$ time curl -XGET '
http://localhost:9200/index_test/_search?q=key:keywords&pretty=true'

^C

real 16m34.575s
user 0m0.006s
sys 0m0.012s

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/f7a0c71d-d54e-4be8-a487-47371ca61456%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(Mark Walkom) #3

4500 shards isn't too bad, we've got 5500 and queries are quick.

You haven't given us much of a clue on what size cluster you have,
including node size, RAM usage, replica count etc, which will impact query
performance. Try installing the elastichq plugin, it has some nice
functionality under Node Diagnostics that lets you see where things might
be getting out of hand.

Regards,
Mark Walkom

Infrastructure Engineer
Campaign Monitor
email: markw@campaignmonitor.com
web: www.campaignmonitor.com

On 30 November 2013 08:38, simonw simon.willnauer@elasticsearch.com wrote:

you are searching on 4500 shards? Is that right? I am just asking there
could be a lot of stuff going on though.

simon
On Friday, November 29, 2013 11:51:35 AM UTC+1, Jason Wee wrote:

Hi, we are experience a strange issue, it happen intermittently. we have
three elasticsearch nodes running version 0.90.7 and started a new cluster.
Then as data coming in, we query one of the nodes, sometime the result
return immediately. Sometime it just hang there forever until ctrl+z is
issued. Not exactly sure what is wrong. Can anyone pinpoint where might be
the possible cause?

[user@node1 ~]$ time curl -XGET 'http://localhost:9200/index_
test/_search?q=key:keywords&pretty=true'
{
"took" : 69,
"timed_out" : false,
"_shards" : {
"total" : 4500,
"successful" : 4500,
"failed" : 0
},
"hits" : {
"total" : 2201,
"max_score" : 6.192957,
"hits" : [ {
"_index" : "index_test",
...
...
...
real 0m0.072s
user 0m0.001s
sys 0m0.003s

[user@node1 ~]$ time curl -XGET 'http://localhost:9200/index_
test/_search?q=key:keywords&pretty=true'

^C

real 16m34.575s
user 0m0.006s
sys 0m0.012s

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/f7a0c71d-d54e-4be8-a487-47371ca61456%40googlegroups.com
.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAEM624agH6H2q-8YuknGzZH_7wWhVB1p9u8kscRRUZO5EOm%2Bzg%40mail.gmail.com.
For more options, visit https://groups.google.com/groups/opt_out.


(Jason Wee) #4

Yes, the index is started with a total of 4500 shards. We are currently
have three dedicated server running for elastic search cluster. The three
nodes has identical hardware configuration. Each node is running intel
i7-4770 cpu with 32GB of ram. Each node is allocated 16GB heap out of 32GB
of total memory and it only run elasticsearch and no other jvm or intensive
cpu processes. Our es cluster currently having 2 replicas and total active
shards currently is 15000.

when curl on http://localhost:9200/_cluster/nodes/node1/stats?all
indices store is around 11GB at this moment. Heap usage hovering in between
6G and 13G. Non heap usage is low at around 60MB. Jvm threads count stay
constant around 120.

when curl on http://localhost:9200/_stats , the path to index current,
indices/index_test/total/indexing/index_current hovering around 240
constant throughout.

Will these information give you any clue what might happen to the
intermittently search slow response?

=> Try installing the elastichq plugin, it has some nice functionality
under Node Diagnostics that lets you see where things might be getting out
of hand.

Thank you, will install this.

/Jason

On Sat, Nov 30, 2013 at 5:52 AM, Mark Walkom markw@campaignmonitor.comwrote:

4500 shards isn't too bad, we've got 5500 and queries are quick.

You haven't given us much of a clue on what size cluster you have,
including node size, RAM usage, replica count etc, which will impact query
performance. Try installing the elastichq plugin, it has some nice
functionality under Node Diagnostics that lets you see where things might
be getting out of hand.

Regards,
Mark Walkom

Infrastructure Engineer
Campaign Monitor
email: markw@campaignmonitor.com
web: www.campaignmonitor.com

On 30 November 2013 08:38, simonw simon.willnauer@elasticsearch.comwrote:

you are searching on 4500 shards? Is that right? I am just asking there
could be a lot of stuff going on though.

simon
On Friday, November 29, 2013 11:51:35 AM UTC+1, Jason Wee wrote:

Hi, we are experience a strange issue, it happen intermittently. we have
three elasticsearch nodes running version 0.90.7 and started a new cluster.
Then as data coming in, we query one of the nodes, sometime the result
return immediately. Sometime it just hang there forever until ctrl+z is
issued. Not exactly sure what is wrong. Can anyone pinpoint where might be
the possible cause?

[user@node1 ~]$ time curl -XGET 'http://localhost:9200/index_
test/_search?q=key:keywords&pretty=true'
{
"took" : 69,
"timed_out" : false,
"_shards" : {
"total" : 4500,
"successful" : 4500,
"failed" : 0
},
"hits" : {
"total" : 2201,
"max_score" : 6.192957,
"hits" : [ {
"_index" : "index_test",
...
...
...
real 0m0.072s
user 0m0.001s
sys 0m0.003s

[user@node1 ~]$ time curl -XGET 'http://localhost:9200/index_
test/_search?q=key:keywords&pretty=true'

^C

real 16m34.575s
user 0m0.006s
sys 0m0.012s

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/f7a0c71d-d54e-4be8-a487-47371ca61456%40googlegroups.com
.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAEM624agH6H2q-8YuknGzZH_7wWhVB1p9u8kscRRUZO5EOm%2Bzg%40mail.gmail.com
.

For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAHO4ityaBRoGoOf9p3GcUNy75eiXtA9F5KK24YwXuLzRZ43txw%40mail.gmail.com.
For more options, visit https://groups.google.com/groups/opt_out.


(system) #5