Learn F5 Technologies, Get Answers & Share Community Solutions Join DevCentral

Filter by:
  • Solution
  • Technology
Clear all filters
Answers

HA with different versions

Hello,
We have a HA pair, both boxes are 11.6 HF6
We would like to upgrade the standby to 12.1.1 and leave the active in 11.6 HF6
Then, we want to switch between them so, 12.1.1 will be active and 11.6 will be Standby
After a week, we will upgrade the other box to.
Is it possible? will it work?

0
Rate this Question
Comments on this Question
Comment made 07-Dec-2016 by Obvious 1

I have the same question. Did you already read this: https://support.f5.com/csp/#/article/K9057

for me it sounds like we need to try it out and only somebody running the exact same versions and who already did it can say for sure if it is working or not.

0

Answers to this Question

placeholder+image
USER ACCEPTED ANSWER & F5 ACCEPTED ANSWER

Running HA with different versions isn't supported. Any issues you're seeking to work around by doing this is going to be outweighed by any issues you experience doing this.

With BigIP you have multiple boot slots. So just upgrade to 12.1 and have 11.6 available in a DR slot. Fallback is as easy as a reboot.

Your other option is to disable HA and just run 2x standalone systems. One at 11.6 and one at 12.1 then write a quick script to enable/disable the VS's and do manual up/down on each unit.

1
Comments on this Answer
Comment made 04-Dec-2016 by NoamRotter 203

Hi Hamish,
Can you provide such a script?

0
Comment made 07-Dec-2016 by brad 375

while it isn't "supported", it has been the process I have followed for quite a few versions including 11.5.4 to 12.1.1 on a non-production HA pair a couple of weeks ago.

As noted, it will not maintain connection tables across the members. Also do not (you cannot) sync the configurations so keep track of changes that are made in the event you need to revert back.

Might want to force the standby system offline.. the downside of that is if there is a failure on the active it won't fail to anything.

I like the suggestions of just upgrading everything and you can always revert by booting to the prior version volume if needed.

Starting to now upgrade my final two, which are the primary production guests.

0
placeholder+image
USER ACCEPTED ANSWER & F5 ACCEPTED ANSWER

The failover will work, but no config synchronization between different cluster members will be possible.

0
Comments on this Answer
Comment made 03-Dec-2016 by NoamRotter 203

Thank you for the quick reply
I just need to have to 12.1.1 to be active.
Changes will not be made during the week.

I suppose that switching back so 11.6 will be the active again will be possible?

This is my plan to revert if needed.
Tomorrow night I will do the upgrade and will update.

0
Comment made 03-Dec-2016 by Niels van Sluis 2742

Switching back will work. But connection- and persistence mirroring information will not be shared between the cluster members. So that could result in a possible interruption of services when switching over.

0
placeholder+image
USER ACCEPTED ANSWER & F5 ACCEPTED ANSWER

We started by doing force offline to the STB machine.
After rebooting the the standby machine to the new partition,
The 2nd machine went disconnected and the 1st machine went disconnected and in standby mode.
This caused network to stop working.
We had to breake the cluster and investigate the root cause of the issue
1st issue was that the 11.6 configuration had warnings and errors which did not comply with with ver.12.1.1
1. We had a LTM policy to forward traffic to a member - we needed to change the member to node since member is not supported.
2. LTM policy to change http to https with this regex was not supported - location "https://[getfield [http::host] \":\" 1] [http::uri]"
After fixing these errors the configuration loaded.
However, we did not have the APM fully working, a message showed us that a another reboot is required.
After reboot, the APM came up correctly.
After that, ASM did not work with error:

Dec 4 18:36:38 BOX2 crit perl[12256]: 01310027:2: ASM subsystem error (asm_start,F5::DbUtils::insert_data_to_table): Row of table PLC.PL_COOKIE_NEGSIG_SIGNATURES is missing sig_id (200001067) -- skipping F5::ImportExportPolicy::Binary
Dec 4 18:36:38 BOX2 crit perl[12256]: 01310027:2: ASM subsystem error (asm_start,F5::DbUtils::insert_data_to_table): Row of table PLC.PL_COOKIE_NEGSIG_SIGNATURES is missing sig_id (200001067) -- skipping F5::ImportExportPolicy::Binary

Running: mysql -u root -p$(perl -MPassCrypt -nle 'print PassCrypt::decrypt_password($_)' /var/db/mysqlpw) -e 'DELETE FROM PLC.PL_COOKIE_NEGSIG_SIGNATURES
Fixed the issue and after running bigstart start asm it was OK.

After that, on the 1st machine (with older version) we did force offline and on the 2nd machine (with the newer ver.) we did release offline
After a few minutes when everyhting start working, we did a complete test and it seems to be working good.

The current situation:
1. No cluster!
2. Active is 12.1.1 - standalone
3. Standby is 11.6 - forced offline

Next steps:
1. Upgrade the other box from 11.6 to 12.1.1 next week
2. Rebuild the cluster


Although the upgrade did not go as planned it looks like the new version is working fine for now.

In future upgrades I thing the conifsync should be disabled to prevent automatic failover and network issues until the manual flipping.

Regards,

0
Comments on this Answer
Comment made 04-Dec-2016 by Niels van Sluis 2742

If you are using HA group, you also want to disable these before upgrading.

0