[
https://issues.apache.org/jira/browse/HDFS-13112?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16355844#comment-16355844
]
Kihwal Lee commented on HDFS-13112:
-----------------------------------
The patch looks good.
- The addition of read locks ensures these edit logging activities do not
collide with edit rolling or HA transitions(In addition to the level of safety
provided by {{noInterruptsLock}}).
- A write lock is not required since these don't change any state other threads
are accessing with a read lock.
And only the secret manager is edit logging with a read lock and all others are
using a write lock, there can be no concurrent edit logging and it covers the
general {{FSEditLog}} thread safety issue, not only the issue between logging
and rolling.
Now, if we believe that it is only unsafe between edit logging and rolling
(i.e. normal edit logging activities are thread safe), we could make
{{getDelegationToken()}}, {{renewDelegationToken()}} and
{{cancelDelegationToken()}} acquire a read lock. And perhaps lease-related
calls too. Any thoughts on this?
In any case, I'm +1 on the patch. If you think we can make additional locking
changes, please file a follow-up jira.
> Token expiration edits may cause log corruption or deadlock
> -----------------------------------------------------------
>
> Key: HDFS-13112
> URL: https://issues.apache.org/jira/browse/HDFS-13112
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: namenode
> Affects Versions: 2.1.0-beta, 0.23.8
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
> Priority: Critical
> Attachments: HDFS-13112.patch
>
>
> HDFS-4477 specifically did not acquire the fsn lock during token cancellation
> based on the belief that edit logs are thread-safe. However, log rolling is
> not thread-safe. Failure to externally synchronize on the fsn lock during a
> roll will cause problems.
> For sync edit logging, it may cause corruption by interspersing edits with
> the end/start segment edits. Async edit logging may encounter a deadlock if
> the log queue overflows. Luckily, losing the race is extremely rare. In ~5
> years, we've never encountered it. However, HDFS-13051 lost the race with
> async edits.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
