DiskThresholdDecider#remain(...) should take shards relocating away into account #8659
Conversation
|
|
||
| DiskThresholdDecider diskThresholdDecider = new DiskThresholdDecider(diskSettings); | ||
| MetaData metaData = MetaData.builder() | ||
| .put(IndexMetaData.builder("test").numberOfShards(1).numberOfReplicas(1)) |
There was a problem hiding this comment.
You put [test][0][p] and [test][1][p], but then create metadata with 1 primary shard and one replica, can you change it here or in the shardSizes map?
|
@martijnvg left a couple of comments |
|
Also, 1.3.x doesn't include the take-relocations-into-account feature, so this may be pretty tough to backport without backporting the original feature, not sure if this is something we should stick in just 1.4.x and up. |
|
@dakrone Good point, didn't realise that. I removed the 1.3.6 version label. I'll update the PR. |
| @@ -223,20 +223,28 @@ public DiskThresholdDecider(Settings settings, NodeSettingsService nodeSettingsS | |||
| /** | |||
| * Returns the size of all shards that are currently being relocated to | |||
| * the node, but may not be finished transfering yet. | |||
| * | |||
| * If includeShardsMovingAway is set then the size of shards moving away is subtracted from the total size | |||
There was a problem hiding this comment.
can we name this subtractShardsMovingAway ? it's clear what it means
|
left one comment - thanks mvg for getting this in!! |
| @@ -223,20 +223,28 @@ public DiskThresholdDecider(Settings settings, NodeSettingsService nodeSettingsS | |||
| /** | |||
| * Returns the size of all shards that are currently being relocated to | |||
| * the node, but may not be finished transfering yet. | |||
| * | |||
| * If subtractShardsMovingAwayRen is set then the size of shards moving away is subtracted from the total size | |||
There was a problem hiding this comment.
subtractShardsMovingAwayRen -> subtractShardsMovingAway
|
LGTM, left one comment about a misspelling |
|
oops, I'll change that. On 26 November 2014 at 09:36, Lee Hinman notifications@github.com wrote:
Met vriendelijke groet, Martijn van Groningen |
…away from a node in order to determine if a shard can remain. By taking this into account we can prevent that we move too many shards away than is necessary. Closes elastic#8538 Closes elastic#8659
martijnvg
force-pushed
the
improvements/disk_threshold_decider_take_shards_relocating_away_into_account
branch
from
f9fc09e
to
c70e09e
Compare
…away from a node in order to determine if a shard can remain. By taking this into account we can prevent that we move too many shards away than is necessary. Closes elastic#8538 Closes elastic#8659
clintongormley
changed the title
martijnvg
deleted the
improvements/disk_threshold_decider_take_shards_relocating_away_into_account
branch
clintongormley
changed the title
…away from a node in order to determine if a shard can remain. By taking this into account we can prevent that we move too many shards away than is necessary. Closes elastic#8538 Closes elastic#8659
lcawl
added
:Distributed/Distributed
Let the disk threshold decider take into account shards moving away from a node in order to determine if a shard can remain.
By taking this into account we can prevent that we move too many shards away than is necessary.
PR for #8538