server-status

Davide Bianchi wrote:
 >
 > Geralyn M Hollerman <gmh2441.TakeThisOut@louisiana.edu> wrote:
  > > I'm trying to understand all the values displayed when browsing to
<font color=green>  > > <a style='text-decoration: underline;' href="http://myserver/server-status;</font" target="_blank">http://myserver/server-status;</font</a>>
 >
 > All the explanation are in the page itself. What's the problem? What's
 > the value you don't understand?

Well, if you must know...

Like I said, I'm trying to understand the values displayed, as the
values I get for my server (Solaris 8, running Apache 2.0.47) don't make
a lot of sense to me. I see the explanations, but they don't seem to fit
with the values I have.

First of all, I don't think I'm understanding the scoreboard correctly -
is it supposed to be a list of workers and their present state, or is
the number of entries set by some parameter? I think the number of
workers I have is set by StartServers * ThreadsPerChild, but I counted,
and there are over 1200 slots (".") in the scoreboard displayed - does
this make sense?

In terms of the parameters, "Srv" - this is supposed to be the "Child
Server number - generation"; the final value I have is "4-0" - so I take
it this means I have 4 "children", and this is the first generation?
Does that correspond to any setting?

Then there's "Acc" - this is supposed to be the "number of accesses this
connection/this child/this slot"; I have values like "5/184/184 and
"0/38/38" - so do I interpret that as a child had 184 (or 3 accesses?
That doesn't make sense.

"SS" is the # of seconds since the beginning of the most recent request
- here, I have values from 0 to 151528 (over 42 days) - my server has
only be "up" for a few days - I can't make sense out of those values.

Finally, "Req", the # of ms required to process the most recent request,
is always 0. I looked at the server status of the server on
httpd.apache.org, and that value is always 0 there, too, so I imagine
it's just too small to be displayed, and that's a good thing. But it
still doesn't make sense that that would ALWAYS be the case - I'd think
you'd get some problem or error sometime.

The reason I asked about "normal" or "average" values was to find out if
there were any "special" values - as in, will "Req" always be "0".

Thanks for any help!

--
Lynn Hollerman.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status 

Geralyn M Hollerman <gmh2441 DeleteThis @louisiana.edu> wrote:
 > the number of entries set by some parameter?

It's the number of "available processes", usually Apache starts in
automatic a number of processes, each one can handle a minimum of X
requests. If the children are not enough (busy site) more are spawned,
so your total number of slots increase because the processes are
destroyed after a certain time.

 > it this means I have 4 "children", and this is the first generation?

Yes.

 > Does that correspond to any setting?

No. It means the child have been generated once and never destroyed.

 > "0/38/38" - so do I interpret that as a child had 184 (or 3 accesses?

The specific child have been accessed 184 times, all in the same slot. As
said, the child have never been destroyed.

 > "SS" is the # of seconds since the beginning of the most recent request

Actually, I think that's a typo and should be milliseconds.

 > is always 0.

In my server isn't 0. I have any value from 1 to over 99999.

 > there were any "special" values - as in, will "Req" always be "0".

Well, if the request is processed just from Apache, I think that '0' is
a good value, if you (like me) use apache as a front-end for other
application servers, then the time required to serve the page strongly
depends on what the application server does.

Davide<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status 

Davide Bianchi wrote:
 >
 > Geralyn M Hollerman <gmh2441.TakeThisOut@louisiana.edu> wrote:
  > > the number of entries set by some parameter?
 >
 > It's the number of "available processes", usually Apache starts in
 > automatic a number of processes, each one can handle a minimum of X
 > requests. If the children are not enough (busy site) more are spawned,
 > so your total number of slots increase because the processes are
 > destroyed after a certain time.

Does that "X" requests correspond to a parameter I've set in httpd.conf?

Does the total # of slots (".") correspond to anything in particular?
How is that determined?

 > No. It means the child have been generated once and never destroyed.

So if I have a number > 0 there does that mean a child has been
recycled?

  > > "0/38/38" - so do I interpret that as a child had 184 (or 3 accesses?
 >
 > The specific child have been accessed 184 times, all in the same slot. As
 > said, the child have never been destroyed.

184 accesses? HOW? The server has only been up 2 days - there are
several other entries in that range as well - there is no way the site
would get that much traffic!

  > > "SS" is the # of seconds since the beginning of the most recent request
 >
 > Actually, I think that's a typo and should be milliseconds.

If that's the case, then yes, that makes more sense.

Is there any particular way that new requests get added to the list on
the server-status page, as in, do they get appended to the bottom or
top, put somewhere at random, what? I can't seem to find any rhyme or
reason to the way mine gets listed.

Thanks again!

--
Lynn Hollerman.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status 

Geralyn M Hollerman <gmh2441.TakeThisOut@louisiana.edu> wrote:
 > Does that "X" requests correspond to a parameter I've set in httpd.conf?

AFAIK the limit is hard-coded in the source.

 > Does the total # of slots (".") correspond to anything in particular?
 > How is that determined?

Every time a new request is received and there are no child
available to serve it, a new process is spawned and a new "slot" is
used. After a certain amount of time if a child is doing nothing the
child is destroyed, also, if the number of MaxRequestPerChild is
reached for a single child, the child is destroyed (this force
the recycling of memory and prevents memory leaks).

 > So if I have a number > 0 there does that mean a child has been
 > recycled?

Basically yes.

 > 184 accesses? HOW?

If a Web page contains 10 images, the client will procede with 10
requests. So it's quite easy to reach a fairly high amount of
requests, even on a single page. That doesn't mean that you have
an high-traffic web site.

 > Is there any particular way that new requests get added to the list on
 > the server-status page

The requests are in child-order.

Davide<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status 

Davide Bianchi wrote:
 >
 > Geralyn M Hollerman <gmh2441.TakeThisOut@louisiana.edu> wrote:
  > > Does the total # of slots (".") correspond to anything in particular?
  > > How is that determined?
 >
 > Every time a new request is received and there are no child
 > available to serve it, a new process is spawned and a new "slot" is
 > used. After a certain amount of time if a child is doing nothing the
 > child is destroyed, also, if the number of MaxRequestPerChild is
 > reached for a single child, the child is destroyed (this force
 > the recycling of memory and prevents memory leaks).

I guess what I'm trying to figure out is what determines the size of the
scoreboard? I have 2 servers, both with roughly the same httpd.conf, but
one has MaxClients set to 500 (ServerLimit set to 20). The scoreboard is
the same size on both, 64 x 20.

And on both, there are 25 entries (the number of threads/child, I
presume), and then "." out to 64; then, another line with 25 more
entries and again, "." out to 64. There are 20 of these lines. What's
the point of this?

  > > So if I have a number > 0 there does that mean a child has been
  > > recycled?
 >
 > Basically yes.

The main thing I'm trying to figure out is why on one server, the
children appear to get recycled (generation is > 0) but on the other
they don't. What controls whether or not a child gets recycled? I know
MaxRequestPerChild is one thing that can destroy a child process, but is
there anything else that would cause a child to be destroyed?

  > > Is there any particular way that new requests get added to the list on
  > > the server-status page
 >
 > The requests are in child-order.

I see that...but then I guess my question should be "what determines
what child gets a particular request"? For instance, my request to view
the server-status page - I notice that always stays with the same child.
But how is it determined just which slot that goes in?

Sorry for so many questions!

--
Lynn Hollerman.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status 

  > > "SS" is the # of seconds since the beginning of the most recent request
 >
 > Actually, I think that's a typo and should be milliseconds.


I have Apache/1.3.12 and SS is showed in seconds elapsed since the
beginning of the previous request (not the current).

For example, when I require a servlet which takes 30 seconds aprox,
the server-status shows successively:

Relative
Event Time time | Srv PID Acc M CPU SS ... Request
1 21:41:27 0 | 7-0 31885 16/450/4230 W 2.10 52 ...
server-status
2 21:41:37 10 | 7-0 31885 17/451/4231 W 2.10 10 ...
myservlet
3 21:41:53 26 | 7-0 31885 17/451/4231 W 2.10 26 ...
myservlet
4 21:42:04 0 | 7-0 31885 18/452/4232 K 2.11 0 ...
myservlet
5 21:42:15 11 | 7-0 31885 18/452/4232 W 2.11 11 ...
server-status

1: server-status is required and it shows 52 SS (52 sec before was the
last requirement on 7-0). Immediately after I required myservlet
2: myservlet is running and 10s before was the last request
(server-status)
3: myservlet is running and 26s before was the last request
(server-status)
4: myservlet has finished and 0s before was the last request
(myservlet)
5: server-status is required, 11s before was the last request
(myservlet)

Important: Look how the columns Relative Time and SS are similar

Hope this helps

By the way, I'm developing a simple monitoring tool for Apache and I
use SS to know if a request has hanged, but as you can imagine, it
doesn´t work when the child hasn't been used for a while.
What I need is to know how long the request has been executing.
Any ideas?<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: server-status