Durability Performance Implications

Author:   Keith Turner
Date:   02 Nov 2016
Reviewer(s)   Josh Elser, Dave Marion

Overview

Accumulo stores recently written data in a sorted in memory map. Before data is added to this map, it’s written to an unsorted write ahead log(WAL). In the case when a tablet server dies, the recently written data is recovered from the WAL.

When data is written to Accumulo the following happens :

Client sends a batch of mutations to a tablet server
Tablet server does the following :
- Writes mutation to tablet servers’ WAL
- Sync or flush tablet servers’ WAL
- Adds mutations to sorted in memory map of each tablet.
- Reports success back to client.

The sync/flush step above moves data written to the WAL from memory to disk. Write ahead logs are stored in HDFS. HDFS supports two ways of forcing data to disk for an open file : hsync and hflush.

HDFS Sync/Flush Details

When hflush is called on a WAL, it does not guarantee data is on disk. It only guarantees that data is in OS buffers on each datanode and on its way to disk. As a result calls to hflush are very fast. If a WAL is replicated to 3 data nodes then data may be lost if all three machines reboot or die. If the datanode process dies, then data loss will not happen because the data was in OS buffers waiting to be written to disk. The machines have to reboot or die for data loss to occur.

In order to avoid data loss in the event of reboot, hsync can be called. This will ensure data is written to disk on all datanodes before returning. When using hsync for the WAL, if Accumulo reports success to a user it means the data is on disk. However hsync is much slower than hflush and the way it’s implemented exacerbates the problem. For example hflush make take 1ms and hsync may take 50ms. This difference will impact writes to Accumulo and can be mitigated in some situations with larger buffers in Accumulo.

HDFS keeps checksum data internally by default. Datanodes store checksum data in a separate file in the local filesystem. This means when hsync is called on a WAL, two files must be synced on each datanode. Syncing two files doubles the time. To make matters even worse, when the two files are synced the local filesystem metadata is also synced. Depending on the local filesystem and its configuration, syncing the metadata may or may not take time. In the worst case, we need to wait for four sync operations at the local filesystem level on each datanode. One thing I am not sure about, is if these sync operations occur in parallel on the replicas on different datanodes. If anyone can answer this question, please let us know on the dev list. The following pointers show where sync occurs in the datanode code.

BlockReceiver.flushOrSync() calls ReplicaOutputStreams.syncDataOut() and ReplicaOutputStreams.syncChecksumOut() when isSync is true.
The methods in ReplicaOutputStreams call FileChannel.force(true) which synchronously flushes data and filesystem metadata.

If files were preallocated (this would avoid syncing local filesystem metadata) and checksums were stored in-line, then 1 sync could be done instead of 4.

Configuring WAL flush/sync in Accumulo 1.6

Accumulo 1.6.0 only supported hsync and this caused performance problems. In order to offer better performance, the option to configure hflush was added in 1.6.1. The tserver.wal.sync.method configuration option was added to support this feature. This was a tablet server wide option that applied to everything written to any table.

Group Commit

Each Accumulo tablet server has a single WAL. When multiple clients send mutations to a tablet server at around the same time, the tablet sever may group all of this into a single WAL operation. It will do this instead of writing and syncing or flushing each client’s mutations to the WAL separately. Doing this increase throughput and lowers average latency for clients.

Configuring WAL flush/sync in Accumulo 1.7+

Accumulo 1.7.0 introduced table.durability, a new per table property for configuring durability. It also stopped using the tserver.wal.sync.method property. The table.durability property has the following four legal values. This property defaults to the most durable option which is sync.

none : Do not write to WAL
log : Write to WAL, but do not sync
flush : Write to WAL and call hflush
sync : Write to WAL and call hsync

If multiple writes arrive at around the same time with different durability settings, then the group commit code will choose the most durable. This can cause one tables settings to slow down writes to another table. Basically, one table that is set to sync can impact the entire system.

In Accumulo 1.6, it was easy to make all writes use hflush because there was only one tserver setting. Getting everything to use flush in 1.7 and later can be a little tricky because by default the Accumulo metadata table is set to use sync. The following shell commands show this. The first command sets table.durability=flush as a system wide default for all tables. However, the metadata table is still set to sync, because it has a per table override for that setting. This override is set when Accumulo is initialized. To get this table to use flush, the per table override must be deleted. After deleting those properties, the metadata tables will inherit the system wide setting.

root@uno> config -s table.durability=flush
root@uno> createtable foo
root@uno foo> config -t foo -f table.durability
-----------+---------------------+----------------------------------------------
SCOPE      | NAME                | VALUE
-----------+---------------------+----------------------------------------------
default    | table.durability .. | sync
system     |    @override ...... | flush
-----------+---------------------+----------------------------------------------
root@uno> config -t accumulo.metadata -f table.durability
-----------+---------------------+----------------------------------------------
SCOPE      | NAME                | VALUE
-----------+---------------------+----------------------------------------------
default    | table.durability .. | sync
system     |    @override ...... | flush
table      |    @override ...... | sync
-----------+---------------------+----------------------------------------------
root@uno> config -t accumulo.metadata -d table.durability
root@uno> config -t accumulo.metadata -f table.durability
-----------+---------------------+----------------------------------------------
SCOPE      | NAME                | VALUE
-----------+---------------------+----------------------------------------------
default    | table.durability .. | sync
system     |    @override ...... | flush
-----------+---------------------+----------------------------------------------

In short, executing the following commands will make all writes use flush (assuming no other tables or namespaces have been specifically set to sync).

config -s table.durability=flush
config -t accumulo.metadata -d table.durability
config -t accumulo.root -d table.durability

Even with these settings adjusted, minor compactions could still force hsync to be called in 1.7.0 and 1.7.1. This was fixed in 1.7.2 and 1.8.0. See the 1.7.2 release notes and ACCUMULO-4112 for more details.

In addition to the per table durability setting, a per batch writer durability setting was also added in 1.7.0. See BatchWriterConfig.setDurability(…). This means any client could potentially cause a hsync operation to occur, even if the system is configured to use hflush.

Improving the situation

The more granular durability settings introduced in 1.7.0 can cause some unexpected problems. ACCUMULO-4146 suggests one possible way to solve these problems with Per-durability write ahead logs.

View all posts in the news archive