Zookeeper, mail # user - Re: Is it possible that follower still hold uncommitted data after truncating and reloading?

Is it possible that follower still hold uncommitted data after truncating and reloading?

anmingyuan 2012-09-05, 04:27

Is it possible that follower still hold uncommitted data after truncating
and reloading?
ZKDatabase.loadDataBase() loads data from the most recent snapshot, but
FileTxnLog.truncate() doesn't delete the snapshot files corresponding to
the deleted log files.
--
With regards,
Mingyuan AN

Re: Is it possible that follower still hold uncommitted data after truncating and reloading?

Mahadev Konar 2012-09-05, 06:15

Mingyuan,
No that should not happen. In case that happens its a bug.

thanks
mahadev

On Tue, Sep 4, 2012 at 9:27 PM, anmingyuan <[EMAIL PROTECTED]> wrote:
> Is it possible that follower still hold uncommitted data after truncating
> and reloading?
> ZKDatabase.loadDataBase() loads data from the most recent snapshot, but
> FileTxnLog.truncate() doesn't delete the snapshot files corresponding to
> the deleted log files.
>
>
> --
> With regards,
> Mingyuan AN

Re: Is it possible that follower still hold uncommitted data after truncating and reloading?

anmingyuan 2012-09-05, 07:24

Mahadev,

Applying https://issues.apache.org/jira/browse/ZOOKEEPER-874 to 3.3.4
fails QuorumPeerMainTest
.testHighestZxidJoinLate.
Seems that truncate logic causes failure.

On Wed, Sep 5, 2012 at 2:15 PM, Mahadev Konar <[EMAIL PROTECTED]>wrote:

> Mingyuan,
> No that should not happen. In case that happens its a bug.
>
> thanks
> mahadev
>
> On Tue, Sep 4, 2012 at 9:27 PM, anmingyuan <[EMAIL PROTECTED]> wrote:
> > Is it possible that follower still hold uncommitted data after truncating
> > and reloading?
> > ZKDatabase.loadDataBase() loads data from the most recent snapshot, but
> > FileTxnLog.truncate() doesn't delete the snapshot files corresponding to
> > the deleted log files.
> >
> >
> > --
> > With regards,
> > Mingyuan AN
>

--
With regards,
Mingyuan AN