[
https://issues.apache.org/jira/browse/HDFS-14861?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16934602#comment-16934602
]
Stephen O'Donnell commented on HDFS-14861:
------------------------------------------
The simplest way to fix this problem may be to increment a counter each time
chooseLowRedundancyBlocks() is called, and reset it each time the iterators are
reset. If they have not been reset for X calls, force a reset:
{code:java}
synchronized List<List<BlockInfo>> chooseLowRedundancyBlocks(
int blocksToProcess) {
final List<List<BlockInfo>> blocksToReconstruct = new ArrayList<>(LEVEL);
int count = 0;
int priority = 0;
for (; count < blocksToProcess && priority < LEVEL; priority++) {
if (priority == QUEUE_WITH_CORRUPT_BLOCKS) {
// do not choose corrupted blocks.
continue;
}
// Go through all blocks that need reconstructions with current priority.
// Set the iterator to the first unprocessed block at this priority level
final Iterator<BlockInfo> i = priorityQueues.get(priority).getBookmark();
final List<BlockInfo> blocks = new LinkedList<>();
blocksToReconstruct.add(blocks);
// Loop through all remaining blocks in the list.
for(; count < blocksToProcess && i.hasNext(); count++) {
blocks.add(i.next());
}
}
callCount++; //>>>> New Counter
if (priority == LEVEL || callCount > threshold) { //>>>> Check counter
against some threshold here
callCount = 0
// Reset all bookmarks because there were no recently added blocks.
for (LightWeightLinkedSet<BlockInfo> q : priorityQueues) {
q.resetBookmark();
}
}
return blocksToReconstruct;
} {code}
If things are working well, then most or all blocks returned by this method
should be scheduled on datanodes, and hence the iterator bookmark should be
close to the head of the list. Resetting it would only cause a few blocks to be
retried.
If things are not working well, then resetting the iterator back to the head of
the list would cause a lot of blocks to be retried and hence it would take
longer to reach the tail of the list. However that would probably indicate
there are problems on the cluster (eg unable to place new replicas, or out of
service replicas).
Provided the time between resets is not too small (eg 30 - 60 minutes) this
would probably be OK.
If blocks are under-replicated (eg from a node failure), skipped blocks are not
a problem - all blocks have to be processed eventually anyway, it does not
really matter what order it happens in, or what is skipped.
However with decommissioning and maintenance mode, a skipped block can prevent
the node from completing the process. Consider decommissioning a few nodes, and
one has relatively few blocks. A skipped block on the smaller node, would cause
it to wait with only a few blocks pending until the other two nodes are fully
processed and the iterator is reset.
> Reset LowRedundancyBlocks Iterator periodically
> -----------------------------------------------
>
> Key: HDFS-14861
> URL: https://issues.apache.org/jira/browse/HDFS-14861
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: namenode
> Affects Versions: 3.3.0
> Reporter: Stephen O'Donnell
> Assignee: Stephen O'Donnell
> Priority: Major
>
> When the namenode needs to schedule blocks for reconstruction, the blocks are
> placed into the neededReconstruction object in the BlockManager. This is an
> instance of LowRedundancyBlocks, which maintains a list of priority queues
> where the blocks are held until they are scheduled for reconstruction /
> replication.
> Every 3 seconds, by default, a number of blocks are retrieved from
> LowRedundancyBlocks. The method
> LowRedundancyBlocks.chooseLowRedundancyBlocks() is used to retrieve the next
> set of blocks using a bookmarked iterator. Each call to this method moves the
> iterator forward. The number of blocks retrieved is governed by the formula:
> number_of_live_nodes * dfs.namenode.replication.work.multiplier.per.iteration
> (default 2)
> Then the namenode attempts to schedule those blocks on datanodes, but each
> datanode has a limit of how many blocks can be queued against it (controlled
> by dfs.namenode.replication.max-streams) so all of the retrieved blocks may
> not be scheduled. There may be other block availability reasons the blocks
> are not scheduled too.
> As the iterator in chooseLowRedundancyBlocks() always moves forward, the
> blocks which were not scheduled are not retried until the end of the queue is
> reached and the iterator is reset.
> If the replication queue is very large (eg several nodes are being
> decommissioned) or if blocks are being continuously added to the replication
> queue (eg nodes decommission using the proposal in HDFS-14854) it may take a
> very long time for the iterator to be reset to the start.
> The result of this, could be a few blocks for a decommissioning or entering
> maintenance mode node getting left behind and it taking many hours or even
> days for them to be retried, and this could stop decommission completing.
> With this Jira, I would like to suggest we reset the iterator after a
> configurable number of calls to chooseLowRedundancyBlocks() so any left
> behind blocks are retried.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
