Some of the mapping missing after ElasticSearch cluster restart


(kakaraft) #1

Hi everybody,

We are facing a problem that, mapping metadata of some types got lost after
ElasticSearch cluster is restarted (got 404 when getting mapping for the
type), but without losing actual data. We have encountered this three times
already, this happened only after upgrading to 0.90.5, but sadly couldn't
find any concrete steps to recreate this yet.

At first we thought data of the type was lost, but after we restore data
for that type (by using a tool to put original data into the specified
type), we saw amount of data in that type was doubled. So the symptom was
mapping lost, while actual data still in ElasticSearch.

So now if we see this again, to manually fix this, we first use PUT
_mapping action, to put back the mapping metadata, so we can access the
type again, then check whether data was lost. But till now, we only see
missing mappings, no data lost.

As so many indices and so many types in the cluster, it's too tedious to
check each type one by one and perform the fix, so trying to find some way
to avoid the mapping lost from happening.

From log file, only see two or three lines of log as below, not other
errors/warnings, not sure whether they are related:

[2013-11-30 23:11:13,537][WARN ][cluster.metadata ] [xweb01.sys2]
failure during [refresh-mapping [log-2013-11-30][[TTS-JP,
Band_K_android]]]java.lang.NullPointerException

Does anyone have any idea what happened there, or facing similar problem?

Basic information of our ElasticSearch cluster:

ES version: 0.90.5
Java version: 1.7.40
OS: 64 bit Linux box
ES nodes: Total 29 nodes, there are 4 master nodes and 20 data nodes
Using local gateway.
Indices are created base on date.

Configuration can be found at: https://gist.github.com/kakaraft/8020408

Thanks & regards.
Ronald

--
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/64575fb6-0463-4b86-b9af-c0f7b3a3e66e%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(Alexander Reelsen) #2

Hey,

do you have a complete stack trace for that NullPointerException in your
logfiles?

--Alex

On Wed, Dec 18, 2013 at 11:58 AM, kakaraft@gmail.com wrote:

Hi everybody,

We are facing a problem that, mapping metadata of some types got lost
after ElasticSearch cluster is restarted (got 404 when getting mapping for
the type), but without losing actual data. We have encountered this three
times already, this happened only after upgrading to 0.90.5, but sadly
couldn't find any concrete steps to recreate this yet.

At first we thought data of the type was lost, but after we restore data
for that type (by using a tool to put original data into the specified
type), we saw amount of data in that type was doubled. So the symptom was
mapping lost, while actual data still in ElasticSearch.

So now if we see this again, to manually fix this, we first use PUT
_mapping action, to put back the mapping metadata, so we can access the
type again, then check whether data was lost. But till now, we only see
missing mappings, no data lost.

As so many indices and so many types in the cluster, it's too tedious to
check each type one by one and perform the fix, so trying to find some way
to avoid the mapping lost from happening.

From log file, only see two or three lines of log as below, not other
errors/warnings, not sure whether they are related:

[2013-11-30 23:11:13,537][WARN ][cluster.metadata ] [xweb01.sys2]
failure during [refresh-mapping [log-2013-11-30][[TTS-JP,
Band_K_android]]]java.lang.NullPointerException

Does anyone have any idea what happened there, or facing similar problem?

Basic information of our ElasticSearch cluster:

ES version: 0.90.5
Java version: 1.7.40
OS: 64 bit Linux box
ES nodes: Total 29 nodes, there are 4 master nodes and 20 data nodes
Using local gateway.
Indices are created base on date.

Configuration can be found at: https://gist.github.com/kakaraft/8020408

Thanks & regards.
Ronald

--
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/64575fb6-0463-4b86-b9af-c0f7b3a3e66e%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/CAGCwEM9hTBykZmSXdRjJhmqdd3tkt_FBcO%3D71-7gBARSukdA7A%40mail.gmail.com.
For more options, visit https://groups.google.com/groups/opt_out.


(system) #3