[
https://issues.apache.org/jira/browse/BOOKKEEPER-304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13398366#comment-13398366
]
Rakesh R commented on BOOKKEEPER-304:
-------------------------------------
@Flavio @Ivan
Thanks for looking the patch. I've few clarifications and please correct me if
I'm wrong.
bq.is that starting the auditor would then be a separate admin task
But I got confused whether to keep the auditor service out of bookie server and
hope you are considering it as an external entity used by admin.
Just my idea is, during the bookie startup only one among them in the bkcluster
will play as auditor(do auditing service) and is part of the bookie server
side. With this approach I feel the recovery mechanism will act quickly on
failures.
Also Auditor recovery feature is a configurable item and can be switched
off/on. What's your thoughts on this?
bq.This patch should use the new LedgerManager. I can use asyncProcessLedgers
to get the list of ledger ids, and with each ID, use readLedgerMetadata to get
the ensembles.
Yeah Ivan, I got the idea of LedgerManager generalization and will refactor the
code.
{quote}.Generating this map should be quite quick (of course we should verify
this statement) and given that bookie crashes should be rather rare, we can
regenerate each time a bookie goes down.
I don't think watchers is a good approach here, as you will end up with as many
watchers as you have ledgers, which could be millions, and in any case, when
the watcher is triggered you have get the whole ledger list again anyhow. Also,
watching is zk specific, so not possible through just the LedgerManager
interface.
{quote}
As I understand, auditor won't maintain a bookie vs ledger index cache. On
demand (failure notification) auditor will start parsing all the ledger
metadata and get suspected ledgers corresponding to the failed bookie(s).
But I'm just thinking the overhead of generating the indexes each time. Will it
be ok?
> Prepare bookie vs ledgers cache and will be used by the Auditor
> ---------------------------------------------------------------
>
> Key: BOOKKEEPER-304
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-304
> Project: Bookkeeper
> Issue Type: Sub-task
> Components: bookkeeper-server
> Reporter: Rakesh R
> Assignee: Rakesh R
> Fix For: 4.2.0
>
> Attachments: BOOKKEEPER-304.1.patch
>
>
> This JIRA discusses how to build bookie -> ledgers cache and this will be
> used by the Auditor to publish the suspected ledgers of failed bookies.
--
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
