#logstash

the RAM probibly get destroyed but zabbix would tell you this

there are memory settings for ES / LS that grab min amounts of memory for the java heap

tbh you really want a bit of RAM free for cache or your all on disk IO / going in to SWAP

we tend to never run servers above 50-60% ram used as the results IO unhappiness - but varies from what you do

Fuwan ask in the #elasticsearch channel, but you are probably hitting your capacity ceiling

Ye, most of the time swap is going over 50%

Will get a seperate machine up for the logging bits, thanks both

swap with ES is bad

you probably want to put it on it's own server

i find SWAP is always bad

fact - we don’t even have a SWAP parition on any server

rather it OOM than chew up the IO on the SAN

fully agree with eper; swap was nice when you actually had problems fitting in more than 2 programs in your memory (ie: less than 512MB ram), but now you'd rather want a program to crash than to run very very slow (and crash anyway in the end)

might be ok if your storage is all SSD based

but even then its not RAM speed :]

it slows it down, so it is bad

hi guys

anyone did an upgrade from 1.1.1 to 1.3.1 ?

elasticsearch that is ..

warkolm: now I've got relocating shards, while still unassigned ones

warkolm: is that normal?

vali: i did recently 1.2.3 to 1.3.0 without any issues

I just did 1.1.1 to 1.3.1 and worked out without any issues also

electrical, thank you for the reply.

vali: np :-)

seems to always upset our API calls when we update :]