Unsolved
This post is more than 5 years old
223 Posts
0
2862
CIFS Server deleted after Datamover reboot
Hello,
today I had to restart both datamover of a NS120 twice.
Each time after the restart the CIFS server was deleted on the NS120 and I had to cretae it new.
I have NS120 Dart (6.0.70.4) and Flare Code (4.30.0.5.525).
Any idea why this happens, or has someone seen this issue on other Systems?
Regards
David
Rainer_EMC
8.6K Posts
0
May 16th, 2013 09:00
No – that shouldn’t happen and I’ve never seen it
Please open a service request so that support can properly troubleshoot
cincystorage
467 Posts
0
May 16th, 2013 10:00
That should never happen. What output did you see from "server_cifs" when the the server was missing?
Druehl1
223 Posts
0
May 16th, 2013 12:00
No, l didn´t see a failover, but the uptime of both datamovers started at 0:00.
Was there something like fastboot on the dm, so that the connection don´t get lost while a rebbot.
I did the reboot in the GUI not by CLI
cincystorage
467 Posts
0
May 16th, 2013 12:00
Agreeded. Did it Failover to the standby datamover?
Rainer_EMC
8.6K Posts
0
May 16th, 2013 12:00
On every system the connection gets dropped when it is rebooted – that is the way TCP/IP works
The client is supposed to re-initiate that connection when the server is available again.
How that looks like to the client depends on the Windows version, the application and what you are doing while the connection isn there
Druehl1
223 Posts
0
May 16th, 2013 12:00
Hi Mark,
the output of the cli showed that the server exists, but the hosts lost the connection to the shares.
´Then I added the CIFS server in the GUI witout joining the domain, and after that it worked again.
I have never seen this before and have no idea what can cause this.
So I will open a case at emc.
Rainer_EMC
8.6K Posts
0
May 16th, 2013 12:00
That sounds differently than the CIFS server being deleted after a reboot ….
cincystorage
467 Posts
0
May 16th, 2013 13:00
Yeah, this sound like normal windows behavior... Windows is stateful, so it will not be happy during reboot - a lot of windows versions don't do this very well.. I suspect reconnecting drives or rebooting the hosts in question would also solve the issue..
This is one the benefits we find using Isilon over celerra... the Windows hosts don't notice...
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
May 16th, 2013 14:00
sure they do, if you are using static smartconnect pool where IP do not failover like they do in dynamic pool ..clients get disconnected.
Rainer_EMC
8.6K Posts
0
May 16th, 2013 15:00
There is no fundamental difference there between VNX and Isilon or other CIFS servers
If the CIFS server reboots or fails over the TCP connection gets dropped as does the SMB session
Both have to be re-established by the client
Whether or not you notice that on client doesn’t change the way it works
What is there difference there with Isilon ?
The way the VNX data mover failover works with always transferring the IP address on failover makes it actually easier on the client than having to look for a new IP
With SMB3 CA it is even better
cincystorage
467 Posts
0
May 16th, 2013 17:00
True Enough, forget about static smartconnect zones..
cincystorage
467 Posts
0
May 16th, 2013 17:00
In my experience with the Celerra platform during failover most windows computers won't reconnect after a data mover panic - they require reconnecting the drive mapping manually.. which usually is easier to tell our desktop users to reboot.
With Isilon and dynamic IP assignment using SmartConnect, the IP and mac move to another node and clients don't even notice a problem. I've successfully don't rolling restarts without any clients noticing mid-day.
christopher_ime
2K Posts
0
May 17th, 2013 00:00
In regards to Isilon, even with dynamic IP allocation (and SmartConnect), SMB connections, or more specifically stateful connections, have some level of disruption. Instead of a change in IP as in the case of static IP allocation (and SmartConnect) when a node goes down and a new IP is assigned upon name zone resolution, even with dynamic IP allocation, when the IP is transferred to another node/interface the MAC changes.
Regardless, bp for SMB connections is static IP assignment mainly because there is no benefit provided by dynamic IP allocation with stateful connections.
Jun_Tan
605 Posts
0
May 17th, 2013 02:00
exactlly, smb3.0 will do better when handling failover operation. the following photo is token from one presetation about SMB3.0 new major features. you can found the presentaiton at here.
cincystorage
467 Posts
0
May 17th, 2013 04:00
We've had cases open during datamover panics (not recently 1.5 - 2 years ago) and they said it was normal for clients to have to reconnect manually because of the delay in moving the mac/ip post-panic.
We no longer use the celerra platform for smb anymore - mostly because we didn't want to keep doing nas migrations every lease cycle hardware refresh.