For backup/restore, do not use cp. There is snapshot/restore for that. It works on primary shards only.
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-snapshots.html You can not reduce shards in an existing index. Use export/import tools and create new index. Jörg On Sun, Oct 26, 2014 at 3:44 AM, Atrus <anhhu...@gmail.com> wrote: > Thanks Mathieu for you response. I will try your suggestions. > > "It's a huge value. Shards can be split between nodes, do you target tu > use 15 nodes?" > > Hi Mat, > > - For examples if I have just one node, shards = 5, replica = 0. Then I > can easily backup the data by "cp /var/lib/elasticsearch/nodename > /somewhere/backup -rfp" > > - Now I add one more node, so the cluster has two nodes, shards = 5, > replica = 0. The shards are redistributed, maybe 1st node holds 0 2 4, 2nd > node holds 1 3. => How can I backup, each node does not hold the whole > data, can not simple cp ... > > - If I update replicas = 1, each node now have full 5 shards, I can easy > cp backup on any node. > > If you know the better way for backup which can handle distributed shards, > plz let me know. > > Thank you. > > PS : Can I reduce shards from 5 to 4 without losing data ? > > On Sunday, October 26, 2014 12:57:40 AM UTC+7, Mathieu Lecarme wrote: >> >> >> >> Le vendredi 24 octobre 2014 10:43:21 UTC+2, Atrus a écrit : >>> >>> - There is 15 shards per index, is this too much or enough ? I've used >>> the default config. I know that this could be effect the load but dont know >>> how to figure out the exact number. >>> >>> It's a huge value. Shards can be split between nodes, do you target tu >> use 15 nodes? >> >> >>> - Is there any way to show the running queries ? something like mysql >>> show process list ? to show what queries have eat CPU alot. I have enable >>> slow log queries >1s but found nothing. >>> >> You can watch HTTP traffic, with pcap (I hack packetbeat, for that). It's >> from the outside, from the inside, use the hot thread. strace can help, too. >> >> >>> - Any suggestion is appreciate. >>> >> Do you poll the _nodes/stat url? a monitoring tool, or a web page like >> kopf? >> >> -- > 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+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/elasticsearch/4ab77f45-177d-4546-b953-5f38c7f4f5d1%40googlegroups.com > <https://groups.google.com/d/msgid/elasticsearch/4ab77f45-177d-4546-b953-5f38c7f4f5d1%40googlegroups.com?utm_medium=email&utm_source=footer> > . > > For more options, visit https://groups.google.com/d/optout. > -- 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+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAKdsXoEAjqyXotjzP3STJcX2vMzjwBtjiWerOe0rF0wOtk1rsg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.