Hi, Nik: Thank you for your practical experience sharing. I''ll remember and follow your advice. Thanks again!
On Saturday, December 28, 2013 3:57:38 AM UTC+8, Nikolas Everett wrote: > > > > > On Fri, Dec 27, 2013 at 4:30 AM, Daniel Guo <[email protected]<javascript:> > > wrote: > >> distributed model >> > > Automatic shard rebalancing works quite well. We're able to do rolling > restarts without losing any redundancy. It is useful to keep in mind that > some things, like scores and suggestions, come from data that is per shard > rather across the whole index. > > >> read-time indexing >> > > I assume you mean real time indexing. That works fine. Our problem is > actually getting the documents built and shipped of to Elasticsearch in a > timely manner, not Elasticsearch being able to ingest them. It is > important to make sure that you have a process for doing on line schema > changes like > http://www.elasticsearch.org/blog/changing-mapping-with-zero-downtime/ . > Those processes can push Elasticsearch to its limit if you do them > multi-threaded/multi-process (shakes fist at PHP). Just don't use so many > threads that you crush Elasticsearch. You'll have to measure that. We > crushed three Elasticsearch nodes with 20 processes but your mileage will > vary. > > >> search performance >> > > So far everything is quite quick and we're happy that we can add more > replicas to increase performance. We're not sure yet if we'll do that. I > suggest setting up whatever kind of performance metrics gathering system > you have in house. Capturing those metrics is pretty simple as you can > just dig them out of the rest api. If you happen to use ganglia feel free > to use our script: > http://git.wikimedia.org/tree/operations%2Fpuppet.git/8509513c2ec7c0114554deac3dbb6aa177ce743a/modules%2Felasticsearch%2Ffiles%2Fganglia > > >> and so on. >> > > As I said before I like the Elasticsearch community. They are helpful. > > Make sure to wait a week to ten days after each release to see if some > critical flaw is discovered. Elasticsearch is pretty well tested but every > other release seems to have had some trouble recently. I doubt this'll > happen every time but you may as well be safe. > > For my use case automatic index creation and automatic field creation more > trouble then helpful. These may be worth turning off for you. They are on > by default because they work well for some significant portion of users and > they make playing around really easy. > > Nik > -- 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 [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/a2f4eede-c4e9-4369-96f0-ee4acfe9afa4%40googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
