Unsolved
This post is more than 5 years old
2 Posts
0
965
Can SyncIQ replication policies affect Oracle database dumps
Good day all,
This is my first time on this site so please forgive the "newbie-ness".
We currently have Linux Admins who create and write Oracle database dumps on to the Isilon NFS mounts (OneFS 7.0.1.4). We then replicate the database dumps over to our "DR" Isilon that resides in a separate data center using SyncIQ 2.5. Occasionally, and this has happened 3 times now, our WAN link goes down thus affecting the SyncIQ (obviously). However, what is puzzling is that when this outage occurs it tends to cause the Linux/Oracle host connection speed to drop from an average of 600Mb/s to 2Mb/s which affects the Oracle dump copy.
Does anyone know if the clients create any sort of connection to the offsite array in order to perform the SyncIQ replication? The way I understand it is the Primary Isilon array connects to the DR Isilon array and if this is the case why would our client connections slow to crawl?
Additional Notes:
We don't utilize Quotas on the Arrays and there is not network throttling between the two sites.
Any insight to this issue would be most appreciated.
Cheers,
Lexotic242
Peter_Sero
1.2K Posts
1
May 11th, 2014 05:00
> Does anyone know if the clients create any sort of connection to the offsite array in order to perform the SyncIQ replication?
No, the client has no idea about SyncIQ and the remote DR site.
> The way I understand it is the Primary Isilon array connects to the DR Isilon array and if this is the case why would our client connections slow to crawl?
Interesting question indeed... How well does the primary Isilon serve other
clients during the WAN outage? How well would it serve the Oracle machine
on a separate share or NFS mount?
Does the Oracle dump overwrite the past dump file,
or does it create a new file? Just to rule out we need
to consider some snapshot issues here.
-- Peter
lexotic242
2 Posts
0
May 28th, 2014 14:00
Hi Peter,
Sorry for the delay, I've been out for the past two weeks.
Good information and I'll work with the Oracle DBA's to ask if the dump files is overwritten or a new one created and if so how many copies are retained.
It's interesting that you mention snapshot issues. We've received the following error:"SyncIQ failed to take snapshot on target cluster. Verify that snapshots are licensed on the target, or disable target snapshots for the policy. Snapshots are unlicensed." but when I've asked support if this is part of the issue they indicate that there is nothing to be concerned about. I'm beginning to think otherwise.
Thank you for your time.
Sandy
mattashton1
93 Posts
0
May 30th, 2014 17:00
Out of curiosity, how big are the database dumps? And are you saving the snapshots?
When you say Linux/Oracle connection, is that from a linux machine running Oracle connecting to the Isilon cluster?
Just trying to clarify the network connections involved here.
As stated before, the Isilon cluster would connect to the DR cluster directly for SyncIQ; no network involvement of any clients. If the WAN link goes down between the Isilon clusters, it will try for a bit, but that should not be affecting any of the local traffic between NFS clients and the Isilon cluster.
Gut reaction says to check routing; if the WAN goes down does the route the packets take to the Isilon change?
traceroute is your friend here.
Cheers,
Matt