Hello,

On Sun, 8 Apr 2007, Melvin Wong wrote:

Tks for replying so fast. Maybe I am not so clear with openafs's concept
and I find the openafs guide not really so updated and clear. I execute
the commands as you has re:
[EMAIL PROTECTED] ~]# vos listaddrs
afs1.ben.muveenet
afs2.ben.muveenet

So. Your second file server is known to your database servers. Fine.

[EMAIL PROTECTED] ~]# vos partinfo afs1.ben.muveenet
Free space on partition /vicepa: 711420 K blocks out of total 734684
[EMAIL PROTECTED] ~]# vos partinfo afs2.ben.muveenet
Free space on partition /vicepa: 948956 K blocks out of total 983080

And both partitions are recognized.

I am trying to attain high availability

There are many different types of availability out there.

You need to know against what problems your infrastructure has to be
immune.

* Failure of power -> UPS
* Failure of one single disk -> Raid 5, 6, 1, 10
* Failure of network -> redundant links
* Failure of entire servers -> Replication
...
* Failure of anything -> infinite number of copies

The more things you want to avoid, the more it costs and the slower it
gets (rule of thumb ;-)

as well as creating a big distributed storage space (hopefully it can
reach 30-40 Tera) with many low-end servers.

Even in this situation it depends on the type of data if AFS is the
right choice. Will it be - let's say - 10 big files each 2-4 Tera or
hundreds of thousands small ones. Can they be grouped in a way or not?
What access patterns do you need. Random or more sequentially?

I've created the 1st afs server with home directory and the users can
login and access their home folder without any problems.
But I'm a bit lost on how should I further expand to afs2, afs3 and so
on.

AFS knows 3 kinds of volumes: read-write (rw), readonly (ro) and backup.
For each volume there is only one(!) rw-instance. It is on one single
server at exactly one partition. So the largest volume you can have
cannot be larger than your largest partition.
You can have many (afak 14) ro-clones of each rw-volume. This ro-clones
can be distributed all over your servers.
And there can be exactly one backup-clone, wich is on the same server as
your rw-volume.

Creating a ro-clone is done by issuing a special vos-command. Cloning
takes time an it is (generally) not recommended to do it too often.
Normally intervals of 24 hours are used, more often is possible, but
increases the load on your servers.

If I create a home directory for my users on afs1, do I need to
create the exact directory on my afs2?

No. You do not need. In fact you can not ;-)

Perhaps there should be a small article on openafs.org, saying what
openafs is and what is not... every 2-4 weeks someone is asking the same
questions. What about the wiki?

Oh there is an FAQ:
http://www.dementia.org/twiki/bin/view/AFSLore/GeneralFAQ


Chris
--
Mitarbeiter an der TU Chemnitz, Fak.f. Informatik, Professur VSR
Buero: Str. d. Nationen 62, B204, D-09111 Chemnitz
Tel: +49 371 531-31118, Fax: +49 371 531-831118
_______________________________________________
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info

Reply via email to