"load average too high" VS "XXX children, max XXX"

I don't understand what is difference when sendmail rejects
connections showing:

1- load average too high
2- XXX children, max XXX

But if load average increases, I think sendmail processes also
increase to reach max children, so it's a consequence of this event.
What do you think ?

Re: "load average too high" VS "XXX children, max XXX"

RICCARDO schrieb:
> I don't understand what is difference when sendmail rejects
> connections showing:
>
> 1- load average too high
> 2- XXX children, max XXX

Two fundamental differences:
- "load average" includes non-Sendmail processes, "children" doesn't
- "load average" counts processes actually wanting to run, "children"
may include sleeping processes

> But if load average increases, I think sendmail processes also
> increase to reach max children, so it's a consequence of this event.

Not necessarily. Each of the two can happen without the other one.
- If you run other things besides Sendmail on the same server,
"load average" will go up but "children" won't.
- If clients connect to your Sendmail service but dont send anything,
"children" will go up but "load average" won't.

HTH
T.

--
Please excuse my bad English/German/French/Greek/Cantonese/Klingon/...

Re: "load average too high" VS "XXX children, max XXX"

On 4 Apr, 12:02, Tilman Schmidt wrote:
> RICCARDO schrieb:
>
> > I don't understand what is difference when sendmail rejects
> > connections showing:
>
> > 1- load average too high
> > 2- XXX children, max XXX
>
> Two fundamental differences:
> - "load average" includes non-Sendmail processes, "children" doesn't
> - "load average" counts processes actually wanting to run, "children"
> may include sleeping processes
>
> > But if load average increases, I think sendmail processes also
> > increase to reach max children, so it's a consequence of this event.
>
> Not necessarily. Each of the two can happen without the other one.
> - If you run other things besides Sendmail on the same server,
> "load average" will go up but "children" won't.
> - If clients connect to your Sendmail service but dont send anything,
> "children" will go up but "load average" won't.
>
> HTH
> T.
>
> --
> Please excuse my bad English/German/French/Greek/Cantonese/Klingon/...

What disadvantages I can find if I increase childrens processes ?
If I add ram module for upgrading from 1 GB to 2 GB can i solve
problem ?
My top command shows:

11:28:43 up 295 days, 1:10, 3 users, load average: 2.77, 2.58,
2.46
174 processes: 173 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait
idle
total 13.4% 0.0% 13.4% 0.0% 0.0% 0.0%
73.0%
Mem: 1032748k av, 1018808k used, 13940k free, 0k shrd,
93724k buff
387792k active, 581324k inactive
Swap: 3068372k av, 254120k used, 2814252k free
348256k cached

Re: "load average too high" VS "XXX children, max XXX"

RICCARDO schrieb:
> What disadvantages I can find if I increase childrens processes ?

Sendmail will run more SMTP sessions in parallel, and as your server is
pretty loaded already, each session will take proportionally longer to
complete.

> If I add ram module for upgrading from 1 GB to 2 GB can i solve
> problem ?

I wouldn't think so. Judging from your "top" output the server is not
paging, just overloaded. Adding RAM might help a little by increased
disk caching, but not very much. Depending on what that machine does,
it might help to add faster disks, a second processor, or just to
offload some of its task to another box.

HTH
T.

--
Please excuse my bad English/German/French/Greek/Cantonese/Klingon/...

Re: "load average too high" VS "XXX children, max XXX"

On 4 Apr, 12:02, Tilman Schmidt wrote:
> RICCARDO schrieb:
>
> > I don't understand what is difference when sendmail rejects
> > connections showing:
>
> > 1- load average too high
> > 2- XXX children, max XXX
>
> Two fundamental differences:
> - "load average" includes non-Sendmail processes, "children" doesn't
> - "load average" counts processes actually wanting to run, "children"
> may include sleeping processes
>
> > But if load average increases, I think sendmail processes also
> > increase to reach max children, so it's a consequence of this event.
>
> Not necessarily. Each of the two can happen without the other one.
> - If you run other things besides Sendmail on the same server,
> "load average" will go up but "children" won't.
> - If clients connect to your Sendmail service but dont send anything,
> "children" will go up but "load average" won't.
>
> HTH
> T.
>
> --
> Please excuse my bad English/German/French/Greek/Cantonese/Klingon/...

I'm not understanding: load average is low, no high cpu usage of
sendmail process, no swap operations.
This children number increases if many MTA connect to my mail server,
I think.
What do you think if I increase MaxDaemonChildren from 150 to 300 ?

Re: "load average too high" VS "XXX children, max XXX"

RICCARDO schrieb:
> I'm not understanding: load average is low, no high cpu usage of
> sendmail process, no swap operations.
> This children number increases if many MTA connect to my mail server,
> I think.
> What do you think if I increase MaxDaemonChildren from 150 to 300 ?

Just try it. If it helps, keep it. If your server starts to swap,
reduce it again. Otherwise it won't hurt.

HTH
T.

--
Please excuse my bad English/German/French/Greek/Cantonese/Klingon/...