3120 Switch Stack Upgrade

[Andrew Hernandez]

Hey,

I've been reading through examples on how to use the stack configuration and how upgrades are supposed to auto-magically upgrade when a member has a different IOS version than than stack master.  Am I doing something wrong in thinking the master should just need the auto-upgrade enabled and maybe the download-sw (path to IOS.bin file) in the config file?

So far I've tried to force the download by the copy-sw command which did not do anything, no logs nor did it try to move it over.  I can reload a switch and see it try to download, errors below.  Is there something I'm missing or just leaving out in the master switch config maybe?

Master Switch IOS version: cbs31x0-universalk9-mz.150-2.SE9.bin

Members IOS version: cbs31x0-universal-mz.122-50.SE3

*Jan  2 19:40:46.951: %IMAGEMGR-6-AUTO_DOWNLOAD_SW_INITIATED: Auto-download-software process initiated for switch numbe
r(s) 2, 3
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW:
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: Upgrading system(s) 2, 3
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: examining image...
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: %Tar checksum error in flash:cbs31x0-universalk9-mz.150-2.SE9.bin
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: %Error opening flash:update/info (No such file or directory)
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: ERROR: Image is not a valid IOS image archive.
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: Software was not downloaded or
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_DOWNLOAD_SW: installed
*Jan  2 19:41:06.161: %IMAGEMGR-6-AUTO_COPY_SW_INITIATED: Auto-copy-software process initiated for switch number(s) 2,
3
*Jan  2 19:41:06.329: %IMAGEMGR-6-AUTO_COPY_SW:
*Jan  2 19:41:06.329: %IMAGEMGR-6-AUTO_COPY_SW: Searching for stack member to act
*Jan  2 19:41:06.329: %IMAGEMGR-6-AUTO_COPY_SW: as software donor...
*Jan  2 19:41:06.329: %IMAGEMGR-6-AUTO_COPY_SW: Software was not copied
*Jan  2 19:41:06.329: %IMAGEMGR-6-AUTO_ADVISE_SW_INITIATED: Auto-advise-software process initiated for switch number(s)
 2, 3
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: Systems with incompatible software
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: have been added to the stack.  The
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: software running on all of the stack
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: members has been scanned, and it has
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: been determined that the stack can be
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: repaired by issuing the following
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: command(s):
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:     archive copy-sw /force-reload /overwrite /dest 1 2
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: Note that this operation
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: will temporarily interrupt
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW: network operation.
*Jan  2 19:41:06.530: %IMAGEMGR-6-AUTO_ADVISE_SW:

Thanks,

Andrew

[Mark Malone · Accepted solution]

Hi

your major and minor numbers need to match for stack to form like below pic

Major Version Number Incompatibility Among Switches

Switches with different major Cisco IOS software versions usually have different stack protocol versions. Switches with different major version numbers are incompatible and cannot exist in the same switch stack.

Minor Version Number Incompatibility Among Switches

Switches with the same major version number but with a different minor version number are considered partially compatible. When connected to a switch stack, a partially compatible switch enters version-mismatch (VM) mode and cannot join the stack as a fully functioning member. The software detects the mismatched software and tries to upgrade (or downgrade) the switch in VM mode with the switch stack image or with a tar file image from the switch stack flash memory. The software uses the automatic upgrade (auto-upgrade) and the automatic advise (auto-advise) features. For more information, see the "Understanding Auto-Upgrade and Auto-Advise" section. 

To see if there are switches in VM mode, use the show switch user EXEC command. The port LEDs on switches in VM mode stay off. Pressing the Mode button does not change the LED mode.

You can use the boot auto-download-sw global configuration command to specify a URL pathname for the master switch to use to get an image in case of version mismatch.

http://www.cisco.com/c/en/us/td/docs/switches/blades/3120/software/release/12-2_40_ex/configuration/guide/3120scg/swstack.html

http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3750x_3560x/software/release/12-2_55_se/configuration/guide/3750xscg/swstack.html#wp1223384

[Andrew Hernandez]

Thanks Mark!

I'm new to switch stacking and your post, and the one I found below, helped me understand the caveats setting up switch stacks.  This morning I gave it another shot and ended up hard coding slot 1 to the master (priority 15) and just plugging in 1 switch at a time.  That helped me see each switch come online and give me opportunity to hard code each to a priority and not let the mac address set which should be master.  Now I can see all 6 switches online. 

Also, I ended up following the document below and used the archive download-sw command on the master and pushing the tar file over.  This helped each switch upgrade auto-magically since it created a directory which was missing prior.  Manually would have worked too, but now I know it's process.

http://www.cisco.com/c/en/us/support/docs/switches/catalyst-3750-series-switches/64898-upgrade-3750-stack.html#maintask1

SW1#sho switch
Switch/Stack Mac Address : 1ce8.5d09.7580
                                           H/W   Current
Switch#  Role   Mac Address     Priority Version  State
----------------------------------------------------------
*1       Master 1ce8.5d09.7580     15     1       Ready               
 2       Member 1ce8.5d09.7180     14     1       Ready               
 3       Member 1ce8.5d09.8180     13     1       Ready               
 4       Member 1ce8.5d09.7a00     12     1       Ready               
 5       Member 1ce8.5d09.5e80     11     1       Ready               
 6       Member 5478.1a0b.7400     10     1       Ready

Switch Ports Model              SW Version            SW Image                 
------ ----- -----              ----------            ----------               
*    1 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M    
     2 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M    
     3 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M    
     4 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M    
     5 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M    
     6 28    WS-CBS3120X-S      15.0(2)SE9            CBS31X0-UNIVERSALK9-M

Now I can look forward to seeing how I can design separate switch stacks within this stack to create an ingress, egress and management network.  If that's possible, not sure, but will find out.

Thanks again.

Andrew

[Mark Malone · Accepted solution]

Yes that looks good now , just one thing even though Cisco say you can stack up to 8/9 depending on device , real world they say its best practice not to go over 6 devices in a stack just as your at that now, this can be due to people getting hit bad with memory leaks after 6 switches  taking down the whole stack, docs dont specify it but if you check out this forum and others you will see issues people ran into when they stacked more saying that i have seen up to 8 run okk t sometimes but the more you stack the bigger the problem will be when it fails

Now I can look forward to seeing how I can design separate switch stacks within this stack to create an ingress, egress and management network.  If that's possible, not sure, but will find out

-----I dont think is possible to have a stack within a stack, what we have done with our 3120s is create a separate vlan for mgmt traffic and then sourced all traffic from that vlan i.e ntp source-int vlan 50 , so prod traffic is separate from mgmt traffic its the best you can do without specific mgmt ports , for extra resiliency you can create a port-channel with 6 interfaces 1 int per blade physically back to your nearest switch for extra throughput as its a stack , so no matter what fails there is still an available path

anyway good luck with it