Guys, I've run into an interesting problem on a 4 node cDOT cluster running 8.3.2P9 today. Basically, I've setup a user account with access to a particular VServer. It's used for creating snapshots and such.
When we do the command:
ssh <vserver> vol show <volume_full_name>
it just hangs, and doesn't show anything. We have to do a Ctrl-C to get it back.
But if I simply do a simple: 'ssh <vserver> vol show' it all works nice and quickly. So wierd. I can even do the show command when there are multiple volumes using a wild card and it works.
ssh <vserver> vol show '<volume_*>'
works, as long as it matches MORE than just the single volume. Using the -instance flag doesn't help either.
I even tried moving the admin lif to another node, thinking that one node had some stuff ssh processes, but it didn't make a differnce.
Thanks, John
Did you try adding -volume volname to the command?
So:
ssh <vserver> vol show -volume <volume_full_name>
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of John Stoffel Sent: Thursday, September 21, 2017 3:09 PM To: Toasters@teaparty.net Subject: ssh <vserver> vol show <volume_full_name> hangs
Guys, I've run into an interesting problem on a 4 node cDOT cluster running 8.3.2P9 today. Basically, I've setup a user account with access to a particular VServer. It's used for creating snapshots and such.
When we do the command:
ssh <vserver> vol show <volume_full_name>
it just hangs, and doesn't show anything. We have to do a Ctrl-C to get it back.
But if I simply do a simple: 'ssh <vserver> vol show' it all works nice and quickly. So wierd. I can even do the show command when there are multiple volumes using a wild card and it works.
ssh <vserver> vol show '<volume_*>'
works, as long as it matches MORE than just the single volume. Using the -instance flag doesn't help either.
I even tried moving the admin lif to another node, thinking that one node had some stuff ssh processes, but it didn't make a differnce.
Thanks, John _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Yup, tried that as well. Really quite strange. I suspect it's a stuck ssh process/thread inside onTap somewhere. But who knows?
Justin> Did you try adding -volume volname to the command? Justin> So:
Justin> ssh <vserver> vol show -volume <volume_full_name>
Justin> -----Original Message----- Justin> From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of John Stoffel Justin> Sent: Thursday, September 21, 2017 3:09 PM Justin> To: Toasters@teaparty.net Justin> Subject: ssh <vserver> vol show <volume_full_name> hangs
Justin> Guys, Justin> I've run into an interesting problem on a 4 node cDOT cluster running Justin> 8.3.2P9 today. Basically, I've setup a user account with access to a particular VServer. It's used for creating snapshots and such.
Justin> When we do the command:
Justin> ssh <vserver> vol show <volume_full_name>
Justin> it just hangs, and doesn't show anything. We have to do a Ctrl-C to get it back.
Justin> But if I simply do a simple: 'ssh <vserver> vol show' it all works nice and quickly. So wierd. I can even do the show command when there are multiple volumes using a wild card and it works.
Justin> ssh <vserver> vol show '<volume_*>'
Justin> works, as long as it matches MORE than just the single volume. Using the -instance flag doesn't help either.
Justin> I even tried moving the admin lif to another node, thinking that one node had some stuff ssh processes, but it didn't make a differnce.
Justin> Thanks, Justin> John Justin> _______________________________________________ Justin> Toasters mailing list Justin> Toasters@teaparty.net Justin> http://www.teaparty.net/mailman/listinfo/toasters
"ssh -vvv ... " should output some more details from client side perspective.
On Thu, Sep 21, 2017 at 9:46 PM, John Stoffel john@stoffel.org wrote:
Yup, tried that as well. Really quite strange. I suspect it's a stuck ssh process/thread inside onTap somewhere. But who knows?
"Momonth" == Momonth momonth@gmail.com writes:
Nope, nothing really shows up beyond a timeout. I even created a new user (and reused the RSA key) on that VServer and it too has the same problem. It just times out eventually. I've got a ticket open with Netapp on this, we'll see what they say.
Momonth> "ssh -vvv ... " should output some more details from client side perspective.
Momonth> On Thu, Sep 21, 2017 at 9:46 PM, John Stoffel john@stoffel.org wrote:
Yup, tried that as well. Really quite strange. I suspect it's a stuck ssh process/thread inside onTap somewhere. But who knows?
Well, just a quick update on my 'ssh <vserver> vol show <vol>' hangs we've been seeing under 8.3.2P9, which is basically nothing. I've been asked if I could try a TO/GB to see if that fixes anything.
At this point, if I'm going to do that, I may as well goto 8.3.2P12 or even jump to 9.1P7 since (amazingly!) the upgrade planning docs are something like 9 and 13 pages for each release. Per-node of course.
Wow... the new upgrade process looks *really* good, assuming that CIFS sessions won't get too hosed.
More when I know more. John
"John" == John Stoffel john@stoffel.org writes:
John> Well, just a quick update on my 'ssh <vserver> vol show <vol>' John> hangs we've been seeing under 8.3.2P9, which is basically John> nothing. I've been asked if I could try a TO/GB to see if that John> fixes anything.
Well, it looks like it's a bug in the Solaris Sparc 11.3 version of SSH somewhere. The command works properly from RHEL 6.6 x86_64 systems, and older Sun Sparc systems.
And it's even a problem with OpenSSH on Solaris 11.3 as well... dammit. I can't make it work reliably with
root@lnd01:~# ssh -V OpenSSH_7.2p2, OpenSSL 1.0.2j 26 Sep 2016
This is after using the command:
# pkg set-mediator -I openssh -V openssh ssh
on the Solaris server to change the defaults to use a different version of ssh... So something on Solaris sucks.
John
-
John Stoffel -
Momonth -
Parisi, Justin