I can't log _single heartbeat interface_ as a bug - but it is a problem.
I can't log _fence himself if disk access is delayed on all nodes for 30
seconds_ as a bug - but it is a problem.
If SAN system is not accessible from ALL Nodes, cluster should wait until at
least 1 node get access, and do not fence until it happened.
I can't log _when SAN swithc reboot, all servers survived easily except OCFS
nodes which all rebooted - it is by design.
Memory leak with file creating is known and fixed.
Main concern is fencing. If I use OCFSv2 for backups (runing between 2 and 3
am at night) and
if OCFSv2 cluster fence himself and reboot at 4pm becaause it lost heartbeat
(even if we have not single open file on OCFSv2 this time) - it is a bug for
me, ebven if it is not a bug for OCFSv2 developers.
Here is a difference, I am concerned about practical design, and I dont like
what I can see. Design is unreliable. I use OCFSv2 to store home
on 1 heartbeat cluster, just to watch how it behave - and it is really
system killer in any failure... IT works now (after few patches)
but I can't count on it as on a reliable application. Reasons - see above (I
can count on Veritas, can count on heartbeat, can count on Cisco pix, and on
may other cluster applications or hardware - but not OCFSv2. Btw, RAC
cluster is not reliable thing as well - it is just good performance
improvement and some short term reliability thing but not a real reliable
cluster. But it is obvious for RAC /too many data are shared/ so not a
surprise.).
I hope to find time (when we finish one big project) and look onto the
modern OCFSv2 version (may be, try to figure out how to fix some of this
problems) but for now, I positioned it in our company as _for non critical
use only_. And I have a few confirmations that I was right.
----- Original Message -----
From: "Sunil Mushran" <[EMAIL PROTECTED]>
To: "Sunil Mushran" <[EMAIL PROTECTED]>
Cc: "Alexei_Roudnev" <[EMAIL PROTECTED]>;
<[email protected]>
Sent: Wednesday, October 10, 2007 6:43 PM
Subject: Re: [Ocfs2-users] Cluster setup
BTW, how many bugs have you logged? The least you can do is just that.
Whining will not get you anywhere.
Sunil Mushran wrote:
Alexei_Roudnev wrote:
Does Oracle tests behavior of OCFSv2 in case of:
- 1,000 different users;
YES. By users I assume you mean processes. 500-1000 per node
on a 8 node cluster.
- host1 appends to the file and host2 truncate it; then host3 rename
file.
YES
- file is removed on node1 but still open on node2;
YES
- one node creates file and other try to rename another file into it.
YES
- 900 GB file sytem and we create 1,000,000 directories with 5,000,000
files in it
No, we limit 32000 files in a directory. Doc in the FAQ.
- directory with 100,000 files inside
No. See above.
- file name length 512 symbols in UTF8 encoding
No, as charactersets are not handled by the fs. But knock yourself out.
- we run 'mkdir x; rmdir x' in a loop for a week...
What so magical about a week. Again, knock yourself out.
etc etc...
??
I am more concerned about OCFSv2 usage as a common file system, and not
so many about LVM + OCFSv2. OCFSv2 looks pretty stable when used for
limited ## of files and limited usage scenario (such as Oracle usage)
but not as a common file system used by thousand of students with
unlimited fantasy...
No, it does not make your morning coffee.
_______________________________________________
Ocfs2-users mailing list
[email protected]
http://oss.oracle.com/mailman/listinfo/ocfs2-users
_______________________________________________
Ocfs2-users mailing list
[email protected]
http://oss.oracle.com/mailman/listinfo/ocfs2-users
