As temporary solution I've bumped PostgreSQL connection limit and
increased spool size to limit despooling count. Now differential backups
run fine going up to 100 (or a little more) idle database connections,
but I'm afraid full backup will fail again.
When job is finished all related idle
Hello.
I've recently switched my Bacula from 5.0.3 to 5.2.6. I'm using
PostgreSQL for storing Bacula database. Everything was working fine on
5.0.3 with my 70GB database. However first night after upgrade director
has died during backup corrupting my database completly. I was about to
clean
But ... xfs is optimized for long streaming reads and writes.
The only reliable alternative for me is ext3 and its performance is
unacceptable when handling directories with lots of small files.
M.
--
Gaining the
Additionally, if the emailserver is using mdir format there are
thousands (millions?) of tiny files and there's a fixed overhead in
opening each file no matter what its size is - that results in slow
speeds when handling lots of small files.
I just sent small test job from my mail server - 15
Out of interest, which filesystem are you using on this volume?
xfs, as always when dealing with lots of small files.
M.
--
Gaining the trust of online customers is vital for the success of any company
that requires
Are you talking about a full backup or some other level? If it is not
a full backup
you will get low speed because bacula only is backing up the changed files
and it takes a long time to go through a filesystem with many thousands of
files
to find out what files have changed.
Both full and
IMHO if Spool Attributes is not enabled, bacula inserts metadata into
the catalog after every file - in your case 660'000 inserts instead of 1
large insert with Spool Attributes.
please enable Spool Attributes and try again.
to prevent shoe-shining of the tape, I would recommend to enable
Hello, what about disk I/O on the mail server? I suppose you use Maildir
format for mail storage, try to mount the file system on the mail server
with noatime option.
Hello.
Yes, I'm using Maildir format. I'm already mounting all volumes with
noatime and nodiratime. Disk I/O seems fine. I
Hello.
I've problem with bacula performance on one of my machines. The machine
in question is my mail server that holds about 220 GB of data in 660
thousands of files. Full backup takes about 2 days and 11 hours with
average speed of 1 MB/sec. All other clients have speeds between 20
did you enable Spool Attributes or Spool Data in the job resource?
did you run the btape tests? do you have compression or encryption
enabled?
your config and some infos about the bacula version may help.
Spool Attributers or Spool Data is not enabled in job resource.
I wasn't running
10 matches
Mail list logo
