×

Warning message

  • Cisco Support Forums is in Read Only mode while the site is being migrated.
  • Cisco Support Forums is in Read Only mode while the site is being migrated.

Is there a limit to number of SNASW links?

Unanswered Question
Sep 22nd, 2005
User Badges:

We have an IBM 3745 with T/R which we disconected and attached a router with SNASW instead. Same macaddress as the 3745. It went fine first but when a lot of people started to logon they hit the wall and the SNASW stopped to respond to new connections. We do not use Virtual T/R, just a plain physicall interface with a SNASW port statement connecting the thing. Is there a limit somewhere?

Regards

Bjorn

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
htranphu Thu, 09/22/2005 - 03:26
User Badges:

Hi Bjorn,


By default, the max number of implicit links on a port is 3000. You can change it with the "max-link" option on the port statement (max=10000). You can check the number of links on the router without listing them all with "show snasw link | i Number".


Regards,

tran

bjr Thu, 09/22/2005 - 04:20
User Badges:

Thank you all.

No I don't think we hit 3000 links. It should be around 600 if one link means one PU, which I think it is.

The router was idling at around 2% CUP and we have 256 Mb RAM, There seemed to be a lot left. We have to research more.


Regards

Bjorn

oerkan Thu, 09/22/2005 - 04:34
User Badges:


Hi Bjorn,


Did you collect any snasw traces such as snasw dlc, ips trace, pdlog at the problem time ?


I would suggest you to open a TAC Service Request for further troubleshooting of this issue by providing all the information you collected at the problem time.


Thanks,


Hakan.



romney Thu, 09/22/2005 - 05:19
User Badges:
  • Cisco Employee,

Since you are using a port tied to a physical interface you are limited to 3000. max-links can only be configured for vtok and vdlc ports. However, as you point out, you weren't anywhere near that number. The first place I suggest you look is in the pdlog (snasw dump pdlog to a tftp server and then look at it with a text editor) to see what type of error messages are being recorded. If that doesn't give enough information to resolve the issue then you will need traces as Hakan suggested. - Ray

bjr Thu, 09/22/2005 - 06:30
User Badges:

Hi again!

No I did(of course) not collect any traces;-). But I dumped the pdlog and had a look at it. Thanks for the tip. Interesting thing. I can see that everthing was runing along fine until everthing started to get sense code 0806 0000. We had a couple of XID problems for one PU but I consider that outside this scope. I checked some of the PU's that got the sense code and they are working fine through the IBM 3745(which we switched back to). Aftre a while it seems that the problem got away and SNASW/VTAM started to accept logons again. We probably need another session to take some traces as well. The users aren't that happy about it though.


Regards

Bjorn

htranphu Thu, 09/22/2005 - 08:45
User Badges:

0806 usually means that the resources (PU,LU...) are not known (active) to VTAM. The same devices may have worked fine via the 3745 because they are defined as different resources under a different switched major node. You can go back and check the VTAM log.

oerkan Thu, 09/22/2005 - 23:31
User Badges:


Hi Bjorn,


If it is not configured yet, I would configure the snasw traces at the next try, specially the dlc trace is very important in this problem.


snasw pdlog exception buffer-size 10000

snasw dlctrace buffer-size 12000 frame-size auto-terse format analyzer

snasw ipstrace buffer-size 5000


Please make sure that you have enough free memory space since buffers are allocated from the main memory.



Thanks,


Hakan,





bjr Mon, 11/28/2005 - 05:26
User Badges:

Hi again!


We tried again this morning without success. It still does not show very much. But a question. If I have more than 3000 links, does it say so somewhere? Or are the new ones just rejetced?


Regards

Bjorn

rclousto Mon, 11/28/2005 - 05:56
User Badges:

From reading the code it looks like you'll get a pdlog message, and it will show on the console if you configure 'snasw pdlog information'.


Regards,

Bob

bjr Mon, 11/28/2005 - 06:32
User Badges:

Ok. The pdlog spans over several pages. Is there anything I can do a search on?


Regards

Bjorn

rclousto Mon, 11/28/2005 - 07:34
User Badges:

Try searching for "Implicit link limit on port reached". There may be a path which could put out "Insufficient storage to activate link" instead.


Regards,

Bob

romney Mon, 11/28/2005 - 06:44
User Badges:
  • Cisco Employee,

Hi Bjorn,


You can issue show snasw link and specify the port name to see just the links using that port (don't be fooled by the Total links count ... that is the total for all links on all ports).


Bob has told you how to spot the situation where you are hitting the max-links limit.


However, based on your previous updates I suspect you have a different problem. Are you still seeing 08060000 in the pdlog? If so, that means there is a problem with the switched major node as Tran pointed out.


You mentioned earlier that the customer is getting frustrated. If that is so then I suggest you open a TAC case and collect the traces that have been recommended. With the correct documentation in hand this should be a very quick resolution.


- Ray

bjr Tue, 11/29/2005 - 00:20
User Badges:

Thanks.


I have a handle on the customer right now so thats cool. No I did not see any 0806 this time. I am sifting through the material right now to see if I find anything. I am also waiting for some more info from the customer to enable me to narrow down the search.

Regards

Bjorn

bjr Mon, 11/28/2005 - 06:39
User Badges:

Hi!


Another thing. I dumped all three logs/traces on to a TFTP server. Dlc and pdlog I can browse. How about the ipstrace? Is that one readable?


Regards

Bjorn

romney Mon, 11/28/2005 - 07:47
User Badges:
  • Cisco Employee,

Hi Bjorn,


no, the ipstrace is in binary format, and we have a tool we use to format it. It would only be needed if the pdlog and dlctrace indicate a problem internal to snaswitch.


- Ray

rclousto Mon, 11/28/2005 - 07:59
User Badges:

The IPS trace is in binary format. We use an internal tool to decode it. It shows the internal process signals between SNASw components and can often be very helpful for Cisco support to look, but you really need the source code and a good understanding of the SNASw internals to make sense out of it.

oerkan Thu, 09/22/2005 - 03:32
User Badges:

Hi Bjorn,


And the memory/cpu usage might be a factor as well.


SNASw consumes 1K per PU and 1.3K for each active LU, and that number goes up to 3.4K when the LU is in session.


What was the memory usage and the cpu load at the problem time ? How many PU/LUs do you have and at what stage you encounter the problem ?


Thanks,


Hakan.

htranphu Thu, 09/22/2005 - 03:46
User Badges:

The "show snasw link" output can vary depending on the IOS release; if "show snasw link | i Number" returns nothing, then try "show snasw link | i number".


Regards,

tran

Actions

This Discussion