Any ideas team? There are only 250 connected users (most threads are dulplicates… e.g John Dow shows up 6 times). The box is a 4-way 2.4 Ghz XEON with 1.5Gb RAM. The index should not be down to 4! It seems to be low 20s and below each time I check. I was not expecting this with such few active sessions. Would enabling transaction logging assist? I know about the sequential vs. non-sequential writes, etc. The negative is that there is only one SCSI/RAID controller. I usuallu like a separate controller for these operations. Would it be worth enabling transaction logs? My consideration is to place on C: with Domino binaries/OS since vs. D: where Dbs reside (on RAID 5 config.). Comments appreciated…
Subject: Availability Index @ 4 & Only 250 users
I have seen the same thing. We called it in to Lotus and it is a known “feature”. There is a technote about changing the ExpansionFactor to make that availability index higher, but it is not a real reading of how your server is performing.
Subject: RE: Availability Index @ 4 & Only 250 users
Thanks… is there a preferred link to the technote? I’ll search on Expansion Factor. I would tend to agree that the # is “off” vs. what it used to mean in R5.
Subject: RE: Availability Index @ 4 & Only 250 users
I’ve been banging my head against the wall about this issue for weeks. I’m in a very similar situation. I’m running a clustered environment with two servers and transaction logging as mail servers with 500 users. Back in November I was at 6.0.2 and the SAI was 90% to 100% most of the time. Upgraded to Domino 6.5 in late November. Now SAI is 0% to low 30%, sometimes higher at night. Transaction logging hasn’t helped my SAI.
Subject: Check notes.ini paramenter server_transinfo_normalize
Check into using this parameter - you may have to tweak it a bit for your environment. Here are the reference numbers from the Lotus KB:
Server Availability Index is Calculating Incorrectly Causing the Domino Server to Become Unavailable ==> # 1103047
Why Does Server Availability Index (SAI) Show a Value of 0 (Zero)? ==> 1086587
There are references to other documentation for clusters that give more detail.
HTH…Scott
Subject: RE: Check notes.ini paramenter server_transinfo_normalize
I’m finding this a very confusing situation, and it seems to be caused by clustering.
I have three R6 servers running three flavours of the code - R6.02, R6.03 and R6.5
None of the servers are in a cluster and all run at an AI value of 100. Even the R6.03 server, which is a PII-266 with 192 Meg RAM running the main SMTP server. The other two servers run Mail and Websites.
Now, we also have have two R6.03 servers in a cluster. They both drop their AI values quite drasically, and the 2nd server always seems to stay a reasonably constant 20 points below the main server. They both peak and trough in the same places.
The second server is purely back-up - no users unless the main server fails. The CPU usage on this box (a dual-xeon Dell Poweredge) is minimal and the Domino tasks aren’t using much of the processor.
I upgraded the server to R6.5.1 and this made no difference to the AI scores.
I removed the 2nd server from the cluster and bang! up it went to AI=100
The main server, now on it’s own in the cluster, is still bouncing between 30 and 70.
The normalize INI setting makes no difference. The AI sometimes even drops to zero, yet the users report no drop in response from the server.
Subject: It seems to be network traffic
Running some more tests points to this being a problem with the way the AI is calculated with network traffic.
Taking the 2nd server out of the cluster reduces trhe traffic from it, and hence the AI goes up.
The main server is still sending and receiving network traffic quit a lot, and the performance graph shows a mirroring between Network Bytes and AI. ie when network traffic goes up, the AI comes down, and vice versa.
The reason users don’t see a slow down is because the network is fast and the server has a 1 gig LAN card.
I think the calculation is overcompensating somewhere.