[
https://issues.apache.org/jira/browse/IMAP-371?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13636371#comment-13636371
]
Eric Charles commented on IMAP-371:
-----------------------------------
Hi Andrzej,
Caching the Mailbox instance in SelectedMailbox makes sense.
I didn't look at the code right now, but definitively this will be a good thing
to do (assuming there is no other place where we read/write to the persistence,
to be checked because if it is the case, the instance will become
desynchronized).
It's fine to add a method on the API (assuming we impact the different
implementation, at least jpa, maildir, inmemory).
If you change something, definitively renaming all MessageManager 'mailbox'
local variables to 'messageManager' + renaming on MailboxManager the
MessageManager getMailbox(MailboxPath mailboxPath, MailboxSession session) to
getMessageManager would be nice.
As usual I will be happy to review and test any patch.
Thx, Eric
> MailboxManager::getMailbox called very often
> --------------------------------------------
>
> Key: IMAP-371
> URL: https://issues.apache.org/jira/browse/IMAP-371
> Project: James Imap
> Issue Type: Improvement
> Components: Processor
> Affects Versions: 0.3
> Reporter: Andrzej Rusin
> Assignee: Eric Charles
>
> MailboxManager::getMailbox(MailboxPath mailboxPath, MailboxSession session)
> is getting called extremely often.
> On my custom RDBMS-based Mailstore implementation it is the biggest hot spot.
> I had to resort to caching the return value of it on Memcache, but even after
> that it uses up around 15% of spent CPU time. It is far too much in my
> opinion. (before I started caching it, fully indexed RDBMS getting of the
> mailbox was taking more than 50% of the global RDBMS load).
> My suspect is the method AbstractMailboxProcessor::addFlagsResponses(final
> ImapSession session, final SelectedMailbox selected, final
> ImapProcessor.Responder responder, boolean useUid).
> It gets the mailbox from MessageManager each time, even if it may be not
> needed.
> This method is called in many CommandProcessors during the
> unsolicitedResponses phase.
> After I am able to really profile and confirm that it's the core of problem,
> I will provide a patch.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
