17.3.5 or 17.6.4 | C9300L-24P-4X-E

[emory.clayton]

Looking for an upgrade suggestion for a C9300L-24P-4X-E. The link below suggest 2 different versions (17.3.5, 17.6.4). They will not be doing anything fancy other than L2 switching (I know, its overkill, its what the clients Administration wants). Other than looking at the affected bugs related to both, and choosing the lesser to 2 evils, what other ways would you go about deciding which one to go with?

https://www.cisco.com/c/en/us/support/docs/switches/catalyst-9300-series-switches/214814-recommended-releases-for-catalyst-9200-9.html

[marce1000 · Accepted solution]

 - I would just use the latest advisory release : https://software.cisco.com/download/home/286322046/type/282046477/release/Bengaluru-17.6.4

 M.

[emory.clayton]

You're suggesting to go with 17.6.4 simply based on the release date?

[Leo Laohoo]

Go with 17.6.4 17.3.5.

Update below.

[balaji.bandi]

Sure i vote for 17.6.4 for stable version

[Leo Laohoo]

If anyone is brave to even consider 16.12.X, please refer to my "documentation" below: 3850 (4 x switches), Firmware version: 16.12.4, Uptime: 1y43w4d

9300, IOS-XE version 16.12.4

3850 stack, 16.12.6

[Jon Porter]

Hi Leo, 

What's your load on these switches and stacked? 

Do you have any 9300s running 17.6.4 or 17.3.6  with the same results? 

Our deployment is moving away from 3650 edge to 9300 (all POE) around 100 already deployed with routing.  

[Leo Laohoo]

Half of our fleet is on 17.3.5 and the other half is on 17.6.4.  Some portion of our 17.6.4 are showing memory leak with the "linux_iosd-imag" process.  This process is all about "telemetry" including "SNMP".  

Just this morning, at 06:00, I had to reboot a stack of two switches (half full) because of the same issue.  That one had an uptime of 25 weeks.  

I will be doing more pro-active reboots in the next two weeks.  

I am not keen on upgrading to 17.6.5 because there are some bugs I am trying to avoid.  One of them is CSCwe54104.

[Jon Porter]

I'm on the same line with upgrading to the latest and greatest as it's always bleeding edge. 

Issue is we are standardizing the IOS everywhere so looking at moving to 17.3.6 code.

Cheers.

Jon

[Leo Laohoo]

Cisco has announced the EoS/EoL for 17.3.X and 17.6.X.  

Everyone is encouraged to start migrating to 17.9.X and 17.12.X. 

[Leo Laohoo]

Stack of 7  x  9300, IOS-XE 17.6.4, Uptime of 13 weeks before I had to perform a reboot of the entire stack on 01 March 2023.  After the reboot, the memory leak continues with the same upward trend.

[dwillia5@highpoint.edu]

Gonna jump in here..even though the thread is a bit dated.  We had the same conundrum, go with the 17.6 train or ride out the 17.3.  We decided to go the way of 17.6, and have encountered an "unpublished" defect in POE with 9200 and 9300 series gear.  After a switch reboots, often we will lose POE to either "some" ports, or all.  Opened a TAC case yesterday after upgrading to 17.6.5 on a handful of brand new 9300 UXM switches...found out about the "unpublished" bug.  Changed trains to 17.3.6 on a few stacks last night, POE working as it should. 

HTH.

[Jon Porter]

Thanks.



I'm looking at 17.6.x code train for the new GRE features so will keep that in mind with POE devices.

[dwillia5@highpoint.edu]

@Jon Porterwe have confirmed with cisco that the unpublished bug only impacts C9300-48H switches.  Whether or not I trust them, that is debatable.  I have witnessed the same symptoms on 9200 series as well, same train of code, 17.6.xxx.