Sync Different

[Aitor_Ibarra]

(If Apple had an HA SAN, maybe that would be their slogan)

This is all 5.8...

I'm having to do a full sync of several HA targets at the moment (my fault - didn't successfully disable Windows Update, so they both rebooted about the same time).

All my targets had to be resynced, and all were set to resync automatically. The actual shut down times of each machine were about a minute apart. Most of my targets had to be forced to resync (full sync), but one didn't. The difference between it and the other targets is that it has no cache, whereas the others have WB cache. Is this why it could start a full sync without me forcing it?

Then, when resync wasn't going as fast as I thought it should, I noticed that one node was not set to use jumbo frames on the sync channel, so I fixed this. I knew this would cause a momentary loss of connection, and I would probably have to start the sync again. Sure enough, I did have to restart the full sync manually, although this time the targets didn't go do down first which was nice. However, that non-cached target continued with its full sync where it left off, with no action necessarry on my part.

So, it seems that non-cached targets have a few advantages when it comes to full sync?

BTW, the ability to throttle sync speed on a per target basis is FANTASTIC, and it's great being able to see an immediate effect via WIndows Resource Monitor on the Total B/second after changing the balance.

[Anatoly (staff)]

(If Apple had an HA SAN, maybe that would be their slogan)

Haha:) Nice one:)

All my targets had to be resynced, and all were set to resync automatically. The actual shut down times of each machine were about a minute apart. Most of my targets had to be forced to resync (full sync), but one didn't. The difference between it and the other targets is that it has no cache, whereas the others have WB cache. Is this why it could start a full sync without me forcing it?

Actualy no, the abscense of cache shouldnt affect automatical synchronization initiation. And the synchronization of HAs should start one-by-one.

Then, when resync wasn't going as fast as I thought it should, I noticed that one node was not set to use jumbo frames on the sync channel, so I fixed this. I knew this would cause a momentary loss of connection, and I would probably have to start the sync again. Sure enough, I did have to restart the full sync manually, although this time the targets didn't go do down first which was nice. However, that non-cached target continued with its full sync where it left off, with no action necessarry on my part.

So, it seems that non-cached targets have a few advantages when it comes to full sync?

Thats interesting actually. So we`ve got two options:
1.Console didnt refreshed
2.The sync continued because data was on safe hard drive (comparing to if it would was in RAM).