Re: [MooseFS-Users] Master failover recovery time

[Alex C. <ac...@in...>]

On 15/11/16 17:08, Gandalf Corvotempesta wrote:
> Hi,
> let's assume a master crash, in a cluster with hundred millions of
> file (many, many terabytes)
> How long does it take for the metalogger to recover and being promoted
> as new master?
> During the whole recovery procedure I/O is blocked?
>
> ------------------------------------------------------------------------------
> _________________________________________
> moosefs-users mailing list
> moo...@li...
> https://lists.sourceforge.net/lists/listinfo/moosefs-users

Metaloggers don't get promoted. Only Follower Masters (Pro only). If you
have a metalogger you can convert it manually to a master or write some
complicated Linux-HA stuff to accomplish this.

LizardFS claims to do this for "free beer" but the HA solution is still
based on services external to the the master server. I will tell you
that LizardFS looked like a great proposition from us until we realised
this. There have also been a few complaints on the LizardFS gitHub about
drops in performance over the last few months that users have said have
not happened in MFS.

In my experience a Master follower takes about 20 seconds max to take
over the MD master role in the Pro version. During that time I/O will
only be blocked to files created since the Master died.

If you have that many files, and need the uptime, again I'd suggest
stumping up the cash. It's way cheaper than anything else we found that
could do the same job.

The only alternatives I can offer you are RozoFS (didn't test much, but
it has very specific topology/number of host requirements, only erasure
coding, so can't split between two nearby Datacernters and guarantee HA
if one DC dies) and BeegFS (fast but again no auto-HA without doing it
yourself - this is planned very soon though). NB BeeGFS asks for a
commercial support license for HA features, ACLs and xattrs, which just
sucks a bit.

Believe me, I've gone through exactly what you have. GlusterFS isn't
worth a penny if you want stabillity, and the big storage guys will
charge you at least high-5 or even 6 figures for even a few TB. EMC
Isilon can't even do active/active between sites, only locally via
Infiniband. This wrote it off for us,

The only other commercial system I'd ask you to look at would be
ExaBlox. We didn't price that but I hear it's "competitive".

Good luck

Alex





--
This message is intended only for the addressee and may contain
confidential information. Unless you are that person, you may not
disclose its contents or use it in any way and are requested to delete
the message along with any attachments and notify us immediately.
This email is not intended to, nor should it be taken to, constitute advice.
The information provided is correct to our knowledge & belief and must not
be used as a substitute for obtaining tax, regulatory, investment, legal or
any other appropriate advice.

"Transact" is operated by Integrated Financial Arrangements Ltd.
29 Clement's Lane, London EC4N 7AE. Tel: (020) 7608 4900 Fax: (020) 7608 5300.
(Registered office: as above; Registered in England and Wales under
number: 3727592). Authorised and regulated by the Financial Conduct
Authority (entered on the Financial Services Register; no. 190856).