[
https://issues.apache.org/jira/browse/BOOKKEEPER-247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13407223#comment-13407223
]
Uma Maheswara Rao G commented on BOOKKEEPER-247:
------------------------------------------------
Hi Ivan,
{quote}
Btw, if you guys want to add changes to the patch, could you clone the github
branch, and make the mods there. It makes it easier to track what the changes
between the patches are.
{quote}
Yeah, I should have done that. Will do from next if any more changes on it.
{quote}
The empty ledger problem im not so sure of how to fix. It's not obvious how to
tell the difference between. There's already a test case for this
#testShouldNotGetAnyFragmentWithEmptyLedger
{quote}
The current fix should solve the single entry problem. But with empty ledger
also we may get the entry as currentEntryId. This will make ledger checker mark
that fragment as underReplicated.
Anyway this fragment will be postponed by worker as lastFragment is in
underReplication state. Once that pendingReplication timedout, this will get
force fenced.
So, with PendingReplications logic, it turns out that, there won't be any
ledger idle more than the pendingReplicationTime out period.
Only my worry is that, In Namenode case,
Just start the namenode and dont write any data. By this time, if any Bookie
goes down from selected quoram. Then after 30sec, if user startes writing then
it will fail with fenced exception because Replication worker already would
have fenced. then this will make one switch unnecessarily.
On startup, keeping system idle for some time may be normal scenario as OM may
start the process one another. To start all processes, it may take some time.
Because of this idleness, it will cause one switch. Another argument is, since
this is startup, one switch should be ok.
> Detection of under replication
> ------------------------------
>
> Key: BOOKKEEPER-247
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-247
> Project: Bookkeeper
> Issue Type: Sub-task
> Components: bookkeeper-client, bookkeeper-server
> Reporter: Ivan Kelly
> Assignee: Ivan Kelly
> Attachments: BOOKKEEPER-247-1.patch, BOOKKEEPER-247.diff,
> BOOKKEEPER-247.diff, BOOKKEEPER-247.patch, BOOKKEEPER-247.patch
>
>
> This JIRA discusses how the bookkeeper system will detect underreplication of
> ledger entries.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
