A new day! A new observation!
When I enable Uplink Balancing, it seems the servers that go dark are in fact random each time - but persistant.
So if server A goes down, even though I disable Uplink Balancing and Server A is accessible again across the Astaro, when I enable UB again, it will be Server A that is unavailable.
This seems to reset after a number of hours -- all of sudden Server A will work fine when UB in enabled and Server C will be unavailable.
If this was happening the other way around it would be an easy guess. I know that UB causes connections to be persistantly chosen using round-robin. If one interface was having issues then random websites would appear to be persistantly unavailable until it reset. However what's happening here is the reverse of that.... which makes absolutely no sense.
Ok if I had to completely hypothesize the ONLY explanation would be:
Uplink Balancing Multipath is applying its persistant round-robin choice to the RETURN traffic of a website request from the Public WAN. This sounds completely off the wall but it would explain why sometimes servers are available and sometimes they are not.
I'm sure there's another reason for it, but I cannot figure this out.
I think its ticket time.
__________________
ASG v7.502 Software -- HP360G5, Quad Xeon E5410, 4GB RAM -- 500 User
|