show memory dead

[RahmaSallm]

One of our cat9500 switches (SVI connection) just stopped working suddenly and we had to hard reboot it. Checked the status of the memory after that and saw:

----------------- show process memory sorted ------------------

Processor Pool Total: 1348707052 Used: 361763832 Free: 986943220
reserve P Pool Total: 102404 Used: 88 Free: 102316
lsmpi_io Pool Total: 6295128 Used: 6294296 Free: 832

PID TTY Allocated Freed Holding Getbufs Retbufs Process
0 0 337668352 74027328 240153304 0 0 *Init*
4 0 79719088 48017168 26118360 0 0 RF Slave Main Th
82 0 7389053584 4852408 18540392 0 0 IOSD ipc task
0 0 0 0 10568592 0 0 *MallocLite*
0 0 114533312376 115452091152 10517656 24595255 1233588 *Dead*
615 0 2258799969152 3624656 6206792 0 0 SISF Switcher Th
503 0 4211368 122208 4131104 849828 0 EEM ED Syslog
261 64 82094680 80417632 2265152 0 0 iosp_dmiauthd_co
247 0 7674608 8323432 1930192 0 0 IP ARP Adjacency
150 0 16272392309208 16199104281920 1852720 0 0 SAMsgThread
525 0 1839592 59904 1810200 0 0 EEM Server
360 0 1727856 936 1780864 0 0 ADJ resolve proc
293 0 29449080 28007328 1405944 0 0 XDR receive
472 0 2217530048 1861669328 1102480 0 0 Crypto CA
1 0 1017448 128872 986192 0 0 Chunk Manager
353 0 1116200 190896 974560 0 0 CEF: IPv4 process

I suspected that there is a memory leak for Dead process sand ran #show memory dead but there are multiple processes under it. Se the attached file. The complete show tech also is attached.

What does this mean? Is it a HW issue or SW issue. We are running 17.6.3. I know it is a bit old but we don't want to upgrade until we know that this is SW issue.

[marce1000]

 - FYI : https://www.cisco.com/c/en/us/support/docs/ios-nx-os-software/ios-software-releases-121-mainline/6507-mallocfail.html#anc28

 M.

[RahmaSallm]

Thank you for that. Does that mean if I suspect a memory leak, I just have to monitor the memory with #show process memory sorted, once a day? Will that help in identifying the process that is causing the memory leak?

[marce1000 · Accepted solution]

    - I would (still) put more effort and upgrading to a latest advisory release of the platform and check if that can improvement

          >... I know it is a bit old but we don't want to upgrade until we know that this is SW issue
  - The thing is that usually the causes are related to 'multiple system interactions' and  as such a single issue is usually not the cause  ; therefore the upgrade-and-check-again methodology is the best way to move forward , 

 M.