Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP

Loading…

Recovery Settings: Change settings (still support old settings) and allow for more dynamic settings #1303

Closed
kimchy opened this Issue · 3 comments

2 participants

@kimchy
Owner

Change the recovery settings names to a better descriptive names (they are not on shard level), from index.shard.recovery to indices.recovery prefix, and allow to change all using cluster update settings API. The settings are: indices.recovery.file_chunk_size, indices.recovery.translog_ops, indices.recovery.translog_size, indices.recovery.compress, indices.recovery.concurrent_streams.

@kimchy kimchy closed this issue from a commit
@kimchy kimchy Recovery Settings: Change settings (still support old settings) and a…
…llow for more dynamic settings, closes #1303.
8ebbd1e
@kimchy kimchy closed this in 8ebbd1e
@aguereca

The page http://www.elasticsearch.org/guide/reference/index-modules/translog.html mention some defaults for index.translog.flush_threshold_size and index.translog.flush_threshold_ops
Those are the same as indices.recovery.translog_ops and indices.recovery.translog_size ??
If so, then why when ES starts I see smaller values : translog_size [100kb], translog_ops [1000]

From my log:
[2011-12-07 14:44:56,098][DEBUG][index.shard.recovery ] [De La Fontaine, Valentina Allegra] using concurrent_streams [5], file_chunk_size [100kb], translog_size [100kb], translog_ops [1000], and compress [true]

@kimchy
Owner

@aguereca which version are you using? it will print it under indices.recovery. The settings you pointed to in translog index module control when its being flushed, while the these settings control how its being recovered.

@aguereca

@kimchy Thanks, I just update to the last version, now is printing under indices.recovery. Thanks for the clarification, now is clear to me the difference on both, I stumble with those settings because I'm trying to track a pesky error: "this writer hit an OutOfMemoryError; cannot commit" that have been happening "randomly" each few days on one of the indexes, and one of the messages that I found in the log after some "failed engine" and before the "Flush failed" was a : "failed to flush shard on translog threshold", so I was wondering is some tweaking of this settings could make the trick.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.