Troubles with Priority Group activations after upgrading from 11.4.1HF4 to 11.4.1HF6

Unfortunately I cannot reformat my question above, it won't let me, so here's the configuration in a better readable format, sorry.

ltm pool lbtest_3306 {
    members {
        server1:mysql {
            address 10.19.1.41
            priority-group 100
            session monitor-enabled
            state up
        }
        server2:mysql {
            address 10.19.1.42
            priority-group 1
            session monitor-enabled
            state up
        }
    }
    min-active-members 1
    monitor tcp
}

I think your load balancing method might be an issue. You need to be using a Ratio LB method for the value to be significant. Try it like this:

ltm pool lbtest_3306 {
    load-balancing-mode dynamic-ratio-member
    members {
        server1:mysql {
            address 10.19.1.41
            priority-group 100
            session monitor-enabled
            state up
        }
        server2:mysql {
            address 10.19.1.42
            priority-group 1
            session monitor-enabled
            state up
        }
    }
    min-active-members 1
    monitor tcp
}

Cheers, Mike

Hi Mike,

thanks for the suggestion - will give this a whirl; but what I do not understand is why the above mentioned configuration works flawlessly on HF4 (and it did that for months). So if this is wrong by design, they probably fixed the error in HF6 too; according to the docs, it shouldn't make a difference, but I'll give it a try and will do some further testings on that.

Thanks!

I think your load balancing method might be an issue. You need to be using a Ratio LB method for the value to be significant.

i do not think so. i think his configuration is correct.

Alex, can you try to set slow-ramp-time to zero (0)?

Nitass,

thanks for the suggestion, I have created a few new pools to play around with and have now activated your suggestion on one of them; will ask the customer to test again, which probably won't be until tomorrow.

Hi everyone,

Mike's suggestion did not change the behaviour. Nitass' suggestion works as expected - I'm just trying to understand why it works with the slow-ramp-time set to 0; at least it does not bounce back and forth anymore, will have to do some more intensive testing over the next days.

Any idea why it's necessary to set the slow-ramp-time when this was not necessary on HF4?

There is a similar issue with v10.2.x, and is fixed in v11.0.0 in the same way nitass suggested. See:

SOL13100: Load balancing may fail when the Slow Ramp Time pool setting is enabled with priority groups ( https://support.f5.com/kb/en-us/solutions/public/13000/100/sol13100 ).

BTW, you must have received a message from F5 that v11.4.1 HF6 had a defect. Won't affect you if apm is not enabled.

I'm just trying to understand why it works with the slow-ramp-time set to 0; at least it does not bounce back and forth anymore, will have to do some more intensive testing over the next days.

it could be a bug. it has been reported by number of customer. you may open a support case to get it fixed.

Thanks for the info; I think they will definitely have to ship a new HF version due to the other bugs in HF6 so let's just hope that this bug has also been fixed in the next version then.