Hbase 18070.branch 2#2683
Merged
Merged
Conversation
…iders, not just default/user-space Pass WALFactory to Replication instead of WALProvider. WALFactory has all WALProviders in it, not just the user-space WALProvider. Do this so ReplicationService has access to all WALProviders in the Server (To be exploited by the follow-on patch in HBASE-25055)
apache#2451) * HBASE-25055 Add ReplicationSource for meta WALs; add enable/disable when hbase:meta assigned to RS Fill in gap left by HBASE-11183 'Timeline Consistent region replicas - Phase 2 design'. HBASE-11183 left off implementing 'async WAL Replication' on the hbase:meta Table; hbase:meta Table could only do Phase 1 Region Replicas reading the primary Regions' hfiles. Here we add 'async WAL Replication' to hbase:meta so Replicas can be more current with the primary's changes. Adds a 'special' ReplicationSource that reads hbase:meta WAL files and replicates all edits to the configured in-cluster endpoint (Defaults to the RegionReadReplicaEndpoint.class -- set hbase.region.replica.catalog.replication to target a different endpoint implementation). Set hbase.region.replica.replication.catalog.enabled to enable async WAL Replication for hbase:meta region replicas. Its off by default. The CatalogReplicationSource for async WAL Replication of hbase:meta does NOT need to keep up WAL offset or a queue of WALs-to-replicate in the replication queue store as is done in other ReplicationSource implementations; the CatalogReplicationSource is for Region Replicas only. General Replication does not replicate hbase:meta. hbase:meta Region Replicas reset on crash of the primary replica so there is no need to 'recover' replication that was running on the crashed server. Because it so different in operation, the CatalogReplicationSource is bolted on to the side of the ReplicationSourceManager. It is lazily instantiated to match the lazy instantiation of the hbase:meta WALProvider, created and started on the open of the first Region of an hbase:meta table. Thereafter it stays up till the process dies, even if all hbase:meta Regions have moved off the server, in case a hbase:meta Region is moved back (Doing this latter simplifies the implementation) hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java Read configuration to see if we need to wait on setting a Region read-enabled (if so, replicas will only flip to enable read after confirming a flush of the primary so they for sure are a replica of a known point) hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/AssignRegionHandler.java If configured, on open of hbase:meta, ask the ReplicationSourceManager to add a ReplicationSource (if it hasn't already). hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/RegionReplicaFlushHandler.java Edit log message. hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/UnassignRegionHandler.java If configured, on close of hbase:meta, update ReplicationSourceManager that a source Region has closed. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceFactory.java javadoc and make constructor private. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceInterface.java Add logPositionAndCleanOldLogs w/ default of the old behavior so CatalogReplicationSource can bypass updating store with WAL position, etc. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java Add creation and start of an CatalogReplicationSource. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceShipper.java Go via ReplicationSource when calling logPostionAndCleanOldLogs so new RS can intercept. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/WALFileLengthProvider.java Javadoc. hbase-server/src/main/java/org/apache/hadoop/hbase/util/ServerRegionReplicaUtil.java Add utility for reading configurations for hbase:meta region replicas. hbase-server/src/main/java/org/apache/hadoop/hbase/wal/WALFactory.java Javadoc. hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestRegionReplicaReplicationEndpoint.java Use define. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/CatalogReplicationSource.java Specical version of ReplicationSource for Region Replicas on hbase:meta. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/CatalogReplicationSourcePeer.java Needs a special peer too (peers are baked into replication though we don't use 'peers' here) hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestMetaRegionReplicaReplicationEndpoint.java hbase-server/src/test/java/org/apache/hadoop/hbase/wal/TestWALProvider.java Tests. Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Huaxiang Sun <huaxiangsun@apache.com>
…eplicationsource warmupRegion called by Master on Region move will instatiate the meta WALProvider as part of its action making it so it is already created by the time we go to open the hbsae:meta Region. Accommodate meta walProvider being already up. hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/AssignRegionHandler.java hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/UnassignRegionHandler.java Pass regionInfo. Needed internally. hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSourceManager.java Add handling if meta wal provider already instantiated when addCatalogReplicationSource runs. hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestMetaRegionReplicaReplicationEndpoint.java Add exercising moving meta around between servers. Test replication keeps working.
…ctionRegistry (apache#2652) Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Huaxiang Sun <huaxiangsun@apache.com>
|
🎊 +1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
… load over meta replica regions
It adds load balance support for meta lookup in AsyncTableRegionLocator.
The existing meta replica mode is renamed as "HedgedRead", client sends scan request to the primary meta replica region first,
if response is not back within a configured amount of time, it will send scan requests to all meta replica regions and
take the first response. On top of the existing mode, a new mode "LoadBalance" is introduced. In this mode, client first
choose a meta replica region to send scan request. If the response is stale, it may send the request to another meta replica region or
the primary region. In this mode, meta scan requests are load balanced across all replica regions with the primary mode as
the ultimate source of truth.
Two new config knobs are added:
1. hbase.locator.meta.replicas.mode
Valid options are "None", "HedgedRead" and "LoadBalance", they are case insensitive. The default mode is "None".
2. hbase.locator.meta.replicas.mode.loadbalance.selector
The load balance alogrithm to select a meta replica to send the requests.
Only org.apache.hadoop.hbase.client.CatalogReplicaLoadBalanceReplicaSimpleSelector.class
is supported for now, which is the default as well. The algorithm works as follows:
a. Clients select a randome meta replica region to send the requests if there is no entry for the location in the stale
location cache.
b. If the location from one meta replica region is stale, a stale entry will be created in the statle location cache
for the region.
c. Clients select the primary meta region if the location is in the stale location cache.
d. The stale location cache entries time out in 3 seconds.
If there is no "hbase.locator.meta.replicas.mode" configured, it will check the config knob "hbase.meta.replicas.use".
If "hbase.meta.replicas.use" is configured, the mode will be set to "HedgedRead".
For branch-2 support, it introduces support for ReversedScan over a specific non-default replica region, this is mainly
for load balance meta scan among all its replica regions.
Remove ConnectionImplementation#setUseMetaReplicas()
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.