Max Servers

[SSuhaib]

Hi DBAs,

Openedge 10.1c, AIX 5.x,

In one of our production database the maximum servers allocated are 100. At any given point in time our concurrent user count does not exceed 120 which includes batch processes. Ours is a GUI application and users does not login in remotely, unless for batch processes.

Please advise if it is safe to reduce this parameter and also what should be the "ideal" number for this parameter.

Thanks in advance.

 

[jdpjamesp]

There's a second parameter that goes hand in hand with that and that's -Ma - clients per server. If -Mn is set to 100 and -Ma is set to 1 you'll be in a bit of a pickle.

 

[Rob Fitzpatrick]

Ours is a GUI application and users does not login in remotely, unless for batch processes.

I don't understand this. If your app is GUI and your back end is AIX, it seems to me your users always log in remotely.

In terms of making a recommendation there is some missing information. You haven't said what resources are available on this box, what other workload runs on it, if any, what these database clients do, etc. For starters, please post the portion of your database log from your most recent startup (starting with the (333) message), about the first 70 lines or so.

 

[SSuhaib]

Thanks Cringer & Rob.

Rob, what I meant is they dont use share memory directly and you are correct it is a remote connection. Below is the copy of .lg file.

[2012/03/08@18:04:25.334+0000] P-70112 T-1 I BROKER 0: (8527) Storage object cache size (-omsize): 1024
[2012/03/08@18:04:25.334+0000] P-70112 T-1 I BROKER 0: (4257) Maximum Number of Clients Per Server (-Ma): 5.
[2012/03/08@18:04:25.334+0000] P-70112 T-1 I BROKER 0: (4258) Maximum Number of Servers (-Mn): 101.
[2012/03/08@18:04:25.334+0000] P-70112 T-1 I BROKER 0: (4259) Minimum Clients Per Server (-Mi): 1.
[2012/03/08@18:04:25.334+0000] P-70112 T-1 I BROKER 0: (4260) Maximum Number of Users (-n): 501.

Here is Promon activity output:

Activity - Sampled at 08/02/12 07:21 for 338:04:40.
Event Total Per Sec Event Total Per Sec
Commits 265571 0.2 Undos 1746 0.0
Record Updates 285973 0.2 Record Reads 2987281615 2454.5
Record Creates 252567 0.2 Record Deletes 38930 0.0
DB Writes 379260 0.3 DB Reads 143783843 118.1
BI Writes 119244 0.1 BI Reads 22539 0.0
AI Writes 87239 0.1
Record Locks 15277584 12.6 Record Waits 47 0.0
Checkpoints 714 0.0 Buffs Flushed 1283 0.0
Rec Lock Waits 0 % BI Buf Waits 0 % AI Buf Waits 0 %
Writes by APW 98 % Writes by BIW 33 % Writes by AIW 51 %
Buffer Hits 98 %
DB Size 9499 MB BI Size 2049 MB AI Size 0 K
FR chain 99 blocks RM chain 10 blocks
Shared Memory 98784K Segments 1
58 Servers, 11 Users (0 Local, 11 Remote, 0 Batch),5 Apws
RETURN - repeat, U - continue uninterrupted, Q - quit:

Again how come -Mn is 101 and promon says 58? Some thing I fail to understand. Anyway, any help is higly appriciated.

 

[Rob Fitzpatrick]

The promon output is somewhat helpful, although it is much mode legible if you put it within [ CODE ] [ /CODE ] tags (without the spaces).

The -Mn parameter is the maximum number of remote servers that can be spawned by the broker. But this only happens on demand, i.e. when a remote client connects; it does not spawn 100 servers when you start the database. The -Mi parameter (minimum clients per server) determines how many clients a server must have before the broker will spawn another server, until all servers have been spawned. Yours is set to the default of 1, which is good. This will provide the best load balancing.

At the time you captured the promon stats there were 58 servers running. This means that from the DB start until this time you didn't have more than 58 concurrent remote (TCP) 4GL clients connected.

Also, the values you see for -Mn and -n will be 1 higher than the numbers you specified as your startup parameter values. That is why you see 101 and 501.

It would be helpful if you provided all of the startup portion of your database log (again, within CODE tags). It looks like there are other settings you may want to change. Also provide some information about your platform, like RAM, and whether AIX and OE RDBMS are 32-bit or 64-bit. Your buffer pool is quite small in relation to your DB size, and your buffer hits (98%) are poor.