because we append edits to the edit logs by default of 300 edits per append.
It would not be wise to copy the files with out shutting down the cluster.
multi reasons not to do this thank memcache flushes/compactions/updates all
working on the data files and directories at the same time.
But as in all other database there if all the tables was marked as read only
and memcache flushes was done and all compaction where done then you MIGHT
be able to copy it. but I would not count on that to be a reliable way to do
backups
Billy
----- Original Message -----
From: "Michael Dagaev"
<[email protected]>
Newsgroups: gmane.comp.java.hadoop.hbase.user
To: <hbase-user-7ArZoLwFLBtd/SJB6HiN2Ni2O/[email protected]>
Sent: Thursday, February 12, 2009 11:15 AM
Subject: Re: Backup again
Erik, thank you for the answer.
I am afraid my question not clear.
I agree that we can backup database
just by scanning it and writing the data somewhere
while the database is running.
However, if we want to backup the database
_by copying the underlying HDFS file system_
we have to stop the database.
That is what I tried to say.
M.
On Thu, Feb 12, 2009 at 7:02 PM, Erik Holstad
<[email protected]> wrote:
Hi Michael!
I don't think that it is true that you can only backup system that are
down,
the backup tools that are posted in HBASE-974 can be used on system that
are
up and running.
What that will do for you is to scan the tables and serialize those to
disk
and the
other way around when importing.
Will that work for you?
Regards Erik
On Thu, Feb 12, 2009 at 2:35 AM, Michael Dagaev
<[email protected]>wrote:
Hi, all
As I understand from the discussions on the list,
we can backup the file system only if Hbase is down.
If we want to run backup when Hbase is up
we can work on the Hbase level, i.e. use scan,
MR export/import utils, etc. but we cannot backup
on the file system level.
Is it correct?
Thank you for your cooperation,
M.
