FG1500D Disk Partitions and HA

[Toshi_Esumi]

Does anyone else had a situation that master and slave HDD2 disk partitions are different? It's a long story how we ended up with this situation while the master has partitioning of 5.6 the slave has partitioning of 6.0, which has additional two partitions. Both are running 5.6.6 now.

I thought it wouldn't even sync in HA but I was proven to be wrong. Now I'm concerning if something goes wrong when we try upgrading this cluster to 6.0.4 again (it once failed due to log disk errors). I have a TAC case open but haven't gotten a reply yet.

[hnmr]

We had once a similar issue with a replacement by RMA, which didn't have the same number of partitions on the disks like our FGT-1500D machine in HA had.

We had to align the disk partitions exactly till it started working (synching).

Please refer to Fortigate HA documentation and look at 

https://www.fortinetguru.com/2016/09/disk-storage-configuration-and-ha/

Please share the resolution (which Fortinet support hopefully will provide you once with)

 BR

Hermann

[hnmr]

and a fresh one:

https://kb.fortinet.com/k....do?externalID=FD45183

[Toshi_Esumi]

We basically checked all these options in the KB. But those are to check the conditions/status but not to fix the partition mismatches. FTNT TAC recommended to run "exe formatlogdisk" on both units. It would probably re-partitioin it back to match 5.6's.

We got a maintenance window for another attempt to upgrade it to 6.0.5 next week. So I'll run this command on the slave then swap them over and run it on the other unit at the beginning of the window. Regardress if this fixes the mismatch or not, I'm planning to upgrade the cluster to 6.0 after making sure the slave at that time has 5.6 partitioning. When it starts upgrading, it upgrades the slave first so at that time the slave will run re-partitioning to match 6.0's. Then after that, they should see matching each other.

I'll update with the outcome next week.

[hnmr]

thanks and a good success and less pains during the update next week!

[Toshi_Esumi]

The command "exe formatlogdisk" didn't do anything good other than formatting log disk partition. Then I proceeded to a regular upgrade process, which re-partitioned the unit that had 5.6's to match 6.0's partitioning. So my conclusion is the partitioning difference is not a part of the conditions to form HA cluster.

Only problem was after the upgrade to 6.0.4, the slave's config (at least checksums) never sync with the master even after a factory-reset on the slave. I have another ticket opened at TAC for that, but that's different topic from this thread.