Forum Discussion

Zdenda's avatar
Zdenda
Icon for Cirrus rankCirrus
Feb 25, 2014

Question about unsucessfull upgade 11.2.1 -> 11.3.0

Hi, last night I tried to upgrade our 3900 LB pair to software version 11.3.0 HF8. Upgrade was done this way:

 

  1. renew license on standby LB, upload new images
  2. install image to unused volume, install hotfix to the same volume
  3. restart and failover

But after failover almost all services were unreachable, then I noticed that pools with standard HTTP monitoring are down, so I changed it to TCP monitor just to make it UP. But services were still not reachable. I got 301 redirect from LB but no response from server side.

 

There was no change to servers that time and strange is that also primary LB unit, eventhough it was not upgraded, showed that servers were down in some pools (application logs checked - nothing).

 

So my questions are: were default HTTP monitorings changed in v11.3.0? Is it possible that LB was not resending HTTP responses to client? I could see connections made to pool members.

 

Thanks, Zdenek

 

design

5 Replies

Sort By
  • Zdenda's avatar
    Zdenda
    Icon for Cirrus rankCirrus
    Feb 25, 2014

    Hi, I made TCPdump on updated unit and it behaves this way:

     

    1. user sends GET request to LB
    2. LB responds with 301 redirect to proper path
    3. user sends new GET request with proper path
    4. LB makes TCP connection to frontend server and sends HTTP Request
    5. LB never get HTTP response only ACK on TCP layer
    6. user never get HTTP response only ACK on TCP layer

    Pool monitors are UP, no ARP issue as I was wondering. ASM was taken out from VIP and anyway not generating anything strange when it was used.

     

    Do you have any clue somebody, or did you experience something like this?

     

    Thanks, Zdenek

     

  • Zdenda's avatar
    Zdenda
    Icon for Cirrus rankCirrus
    Mar 03, 2014

    Hello, just trying to get it visibe. Did somebody experience the same or similar issue? Otherwise I create support ticket.

     

    Thank you, Zdenek

     

  • JG's avatar
    JG
    Icon for Cumulonimbus rankCumulonimbus
    Mar 04, 2014

    What was the status of the active and standby at the time, both "active"?

     

    Why did you do "11.2.1 -> 11.3.0" anyway? 11.4.1 would be a much worthier version (and less buggy than the base hotfix-less v11.3) to upgrade to.

     

  • Zdenda's avatar
    Zdenda
    Icon for Cirrus rankCirrus
    Mar 04, 2014

    Failover was running correctly, there was only one LB unit active in one time. So after failover the "non-primary" unit was active and handling all the traffic.

     

    We choosed version 11.3.0 HF8 because we are using it in test environment and we run this version in couple of another LB cluster in another environments. I am not sure about 11.4.1, it's new software and we dont have tested it yet..