HDFS-17685: Option to explicitly choose DFS client lease renewal interval #7215
+123
−18
Conversation
This file contains 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
|
💔 -1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
ndimiduk
approved these changes
Dec 9, 2024
| <value>0</value> | ||
| <description> | ||
| If set between 0 and 30000 inclusive, HDFS clients will renew leases for files they are writing at this interval. | ||
| If dfs.client.lease.renewal.interval.ms is not set and ipc.client.rpc-timeout.ms is set between 0 and 60000, |
There was a problem hiding this comment.
What is magic about the 60000 maximum value of ipc.client.rpc-timeout.ms ? I guess that comes from existing behavior?
There was a problem hiding this comment.
It comes from existing behavior, yes. The lease renewal interval isn't allowed to be longer than half of the NameNode's timeout, which is hard-coded at 60 seconds.
zeekling
suggested changes
Dec 11, 2024
| return min; | ||
| } | ||
|
|
||
| for (DFSClient c : dfsClients) { |
There was a problem hiding this comment.
Yes, but I think it would make the code more complicated. The logic I want is that the lowest valid value from getLeaseRenewalIntervalMs() takes precedence over any value from getHdfsTimeout(). So, doing that in one loop would require tracking two mins.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Description of PR
Currently, DFSClients send lease renewals to the NameNode at an interval equal to half of {{ipc.client.rpc-timeout.ms}}. This logic dates back to 2009 in HDFS-278. At my company, we are interested in using short DFS client timeouts (< 10 seconds). However, we're currently hesitant to do so because that would flood the NameNode with lease renewals.
I propose a setting {{dfs.client.lease.renewal.interval.ms}} that, if nonzero, would be used as the lease renewal interval instead of {{ipc.client.rpc-timeout.ms / 2}}. This would be useful for advanced users that want to control their RPC timeout and lease renewals separately.
How was this patch tested?
Unit tests are included in this PR for the new logic. I have also tested the new feature in my company's HBase infrastructure.
For code changes:
Object storage: have the integration tests been executed and the endpoint declared according to the connector-specific documentation?not applicableIf adding new dependencies to the code, are these dependencies licensed in a way that is compatible for inclusion under ASF 2.0?not applicableIf applicable, have you updated thenot applicableLICENSE,LICENSE-binary,NOTICE-binaryfiles?