toasters September 2012

toasters@lists.teaparty.net

26 participants
22 discussions

ICMP Fragmentation Needed requests not being honored?
by Ray Van Dolson 26 Sep '12

26 Sep '12

Troubleshooting an issue with SMTP mail delivery for autosupport on an IBM N6240 (FAS 3050 I believe). Mail conversation starts up but then never succeeds fully. Mail server just shows a reset or that the NetApp dropped connection. Ran pktt during the SMTP conversation on the NetApp side and saw that we were receiving ICMP Fragmentation Needed messages (requesting MTU 1400) from an upstream router. However, the NetApp continued to send its data at full frame size (1500 bytes - overhead or 1448 bytes). These don't get through and the SMTP conversation stalls. Confirmed that delivery via a local SMTP server which doesn't require fragmentation works fine. This is ONTAP 8.0.2P3. Anyone have any idea why the network stack isn't honoring the fragmentation requests? Thanks, Ray

1 1

NMC on Mac?
by Christopher S Eno 25 Sep '12

25 Sep '12

Hi, Anyone tried to run the Linux version of the NetApp management console to run on a Mac? NetApp/DFM Guys: any chance of a native Mac version of the NMC? Even if unsupported like system manager for Mac? ---- Scott Eno s.eno(a)me.com

2 2

Re: e0M port as default gateway?
by Ray Van Dolson 25 Sep '12

25 Sep '12

On Thu, Aug 09, 2012 at 10:54:57PM +0000, Jayanathan, David wrote: > Hi Ray, > > We allow all management activity to go through the same interfaces > that data goes through. Only on our Cluster Mode implementations do > we use e0M for management ports, otherwise we only plug in e0M in > order to have the RLM/SP connections on the network. > > Hope this helps! > > -David Thanks, David. We are running cluster mode on these systems (two controllers). Key thing we're seeing is that NFS/CIFS file serving activity grinds to a halt when we lose connectivity to AD/NIS servers (the communication for which currently goes through e0M due to default gateway setup). I presume this is expected behavior, but am a bit surprised that NFS file serving would hang for activity not involving an NIS UID. Presumably any NFS requests gets mapped to a UID and would require an NIS lookup? Anyone run into this before? Ray > > -----Original Message----- > From: toasters-bounces(a)teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Ray Van Dolson > Sent: Thursday, August 09, 2012 2:26 PM > To: toasters(a)teaparty.net > Subject: e0M port as default gateway? > > Hello all; > > We have some N6240's (rebadged FAS3xxx series) on which the default > gateway is set to an IP that routes out the e0M interface. > > As such, if the e0M port (wrench port) goes offline, essentially > functionality on that controller ceases as communication with AD, > NIS, etc. stops working. > > We have 10GbE ports on these devices -- we could easily configure the > default gateways to use one of those instead, but I am unclear as to > what best practice is here. We access management functionality > through the IP bound on the e0M port now. > > How do the rest of you set things up? > > Ray

4 4

OnCommand Insight Report
by Steffen Knauf 25 Sep '12

25 Sep '12

Hello, i installed OnCommand Insight Report and create some reports in Cognos. It is possible to create a report with multiple Querries (in Cognos)? I don't find a way to merge my custom querries into one report. I have one report for every Querry at the moment. Any documentation or tips are welcome! greets Steffen

1 0

trying to upgrade a netapp simulator in anticipation of upgrading tungsten
by Randy Rue 25 Sep '12

25 Sep '12

Hey there, We're running a NetApp simulator running 7.3.5 and have connected iSCSI clients that we've upgraded SnapDrive on to 6.4. Now we're ready to upgrade the simulator to 8.1.1. I was under the impression that upgrading a simulator meant shutting down the simulator and running the new version's simultator installer, where it would sense the previous install and upgrade it in place. But a look at the NOW site shows only a VMware appliance for 8.x, looks like the approach has changed? Is there any way to upgrade a 7.3.5 simulator in place to 8.1.1, preserving existing LUNs? Or to migrate or copy existing LUNs from a 7.3.5 simulator to an 8.1.1? Randy

3 3

Upgrading SnapDrive without reconnecting LUNs?
by Randy Rue 22 Sep '12

22 Sep '12

Hello All, We're preparing to upgrade a v3170 running 7.3.5 to 8.1.1 and will first need to upgrade a number of SnapDrive clients to 6.4 or higher. Our DBA's are testing the SnapDrive upgrade and so far have been unable to do so without needing to recreate all their connections (essentially the iSCSI disks mounted in the SQL cluster go away and they need to reconnect to the LUNs). Is there a better way to do this? Seems like in a perfect world, we'd just run some .exe to upgrade SnapDrive, reboot and find all as it was? Hope to hear from you, Randy

5 5

limit concurrent de-dupe processes?
by Randy Rue 18 Sep '12

18 Sep '12

Hello All, I'm searching through previous stuff looking for this but hoping someone will know it offhand. We're struggling with a v3170 running 7.3.5 that has a number of volumes being deduped automatically when their changed blocks hit 20%. You may know that the SIS process is a good example of a process that relies heavily on the Kahuna namespace and tends to hit one CPU hard while everything else stumbles. Do I remember correctly that there's an option somewhere that limits the maximum number of concurrent SIS processes? Or do I need to script something weird. Randy

2 2

CIFS with Unix sec. style and recent MacOS 10 and MS Word
by Steve Losen 11 Sep '12

11 Sep '12

Hi Toasters, We are sharing a Unix sec. style qtree to (at least one) MacOS 10.7 client via CIFS and they are having problems saving files with MS Word. The client writes the new file to a temporary file and then fails to rename it into place. The temp file exists and is correct because we can verify this from a NFS client. A Unix sec. style share appears to be a VFAT (not NTFS) volume to a CIFS client. Under the hood the filer enforces Unix permissions, of course, but all that is hidden from the CIFS client. I suspect that MacOS is looking for a permission that VFAT does not support and is erroneously assuming it cannot rename the file. Just curious if this rings a bell with anyone and if there is any documentation or workaround for it. We are running DOT 8.0.1 7mode on the filer. I need to get together with the user who is having trouble and use the sectrace command to see if the filer is actually denying access. Steve Losen scl(a)virginia.edu phone: 434-924-0640 University of Virginia ITC Unix Support

3 2

Getting actual file paths from lock status
by Dan Finn 07 Sep '12

07 Sep '12

We are having an issue on 2 linux servers that we think may be related to locking. Lock status shows many locks that are GWAITING, like so: NLM[crater.pp.bcinfra.net,13260]: 94691143974912:18446649382565576704 1 GRANTED (0xffffff02c3600e00) NLM[carlsbad.pp.bcinfra.net,679]: 71133248356352:18446672940461195264 1 GWAITING (0xffffff02aa227428) NLM[carlsbad.pp.bcinfra.net,682]: 71107478552576:18446672966230999040 1 GWAITING (0xffffff0423515230) NLM[carlsbad.pp.bcinfra.net,684]: 71171903062016:18446672901806489600 1 GWAITING (0xffffff04cfccaa10) NLM[carlsbad.pp.bcinfra.net,685]: 71167608094720:18446672906101456896 1 GWAITING (0xffffff03bc421818) NLM[carlsbad.pp.bcinfra.net,687]: 71159018160128:18446672914691391488 1 GWAITING (0xffffff049a4a5e00) NLM[carlsbad.pp.bcinfra.net,689]: 71116068487168:18446672957641064448 1 GWAITING (0xffffff02a966d038) NLM[carlsbad.pp.bcinfra.net,693]: 71047349010432:18446673026360541184 1 GWAITING (0xffffff0332284620) We are thinking we may be running into the same issue as this person: http://comments.gmane.org/gmane.linux.nfs/42929 but we are already doing NFS over TCP. I believe that the hex looking number from lock status (ie 0xffffff0332284620) is possibly an inode. Anyone know if that's true? If so, how can I take that and get an actual file name? Thanks, Dan

3 2

ipspaces vs static net routes
by Fletcher Cocquyt 07 Sep '12

07 Sep '12

Hi we are running 8.1 on our 3270 clusters with 17 vfilers on 5 different vlans The vfilers currently share the same default IPspace and we are (understandably) seeing routing issues (asymmetric routing) when NFS clients outside the vFiler's vlan attempt to communicate with the vFiler. (these clients are the exception > 95% of clients are in the same routable net and don't have any issues) What are the opinions about adding new IPspaces vs just adding 5 static net routes to the shared routing table so routing works? thanks, Fletcher

2 2

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

toasters September 2012