On 20.01.2017 03:59, smith wrote:
Hi, chris:
"busy" is the same as "active".
When not use <Executor>, our busy thread always keep under 10 while the
currentThreadCount keeps high (these are get from tomcat manager), So we really don't
know how many threads are truly free. How many are in keepAlivedStatus
That may be my fault, in my contributions earlier.
I believe that it is reasonable to assume that "busy" also counts the threads which are
currently "in keepAlive" status. That is because these threads are not "free" to handle
new requests yet from other clients than the one which "owns" the connection on which they
are in keepAlive.
But I do not know the code, so someone else may want to confirm/correct this.
Anyway, you do have a simple way to verify this : reduce the keepAlive to 5 seconds,
instead of 20, and see how it impacts the counts that you are seeing.
If keepAlive counts as "busy", then you should see less "busy" threads on
average.
If keepAlive does not count as busy, then you will see no difference.
One more proviso : the above is for the case where clients connect to tomcat directly,
without a front-end, which may itself create a pool of permanent connections to tomcat.
(Chris already mentioned that).
If there is such a front-end, and it does something like "pings" on each connection to the
tomcat back-end connections all the time, then you would have, on the tomcat side, an
alsmost constant number of threads in keepAlive wait status.
All in all, considering your ultimate purpose (to determine when your tomcat is getting
close to the limit of what it can handle, and decide if you need more tomcats or not), I
believe that your current focus on the number of tomcat threads active or not, may not be
the best diagnostic tool.
Maybe a better tool would be to look at the number of connections which are waiting to be
accepted. This should always be 0, or small. If it starts growing, then that is a clear
signal that tomcat is getting saturated. (This is related to the "acceptCount" attribute
of the Connectors).
Another tool may be to look at how long it takes tomcat to respond to requests (or at
least some specific types of requests). You can probably set a baseline, when you know
that tomcat is really not heavily loaded (maybe on your development system e.g.). And
then, when you see this response time increasing seriously, is when you should start
analysing why. (*)
This kind of approach would have the advantage of focusing on what really matters in the
end : the ability of your server (as a whole : hardware, OS, tomcat and applications
included) to handle a given number of requests within a given time interval, with /your/
mix of clients and applications.
(*) because the reason may be for example that you need more memory, not more
threads.
-----Original Message-----
From: Christopher Schultz [mailto:ch...@christopherschultz.net]
Sent: Thursday, January 19, 2017 4:38 PM
To: Tomcat Users List
Subject: Re: FW: tomcat 8080 thread not reduced
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
Smith,
On 1/18/17 8:25 PM, smith wrote:
I don't care if the threads will be reduced, I just want to know why.
Okay.
And we want to use the account to determine when the tomcat capacity
is not enough that we need to add max configuration or add new tomcat
servers.
Set your initial and max threads to the same value (pool size =
constant) and then monitor the "active count" with a Nagios warning at e.g. 80%
usage.
Since not use the <Executor>, the busy thread account also cannot tell
us the correct active threads count.
"busy" is the same as "active".
In another email thread, you said if use <Executor>, it will tell us
the right active thread count not just busy count, right?
I would always use an <Executor> for at least two different reasons:
1. Thread management (e.g. reducing threads if necessary) 2. Shared
thread-pools (no need to have port 8080 and 8443 with separate pools)
- -chris
-----Original Message----- From: Christopher Schultz
[mailto:ch...@christopherschultz.net] Sent: Wednesday, January 18,
2017 3:28 PM To: Tomcat Users List Subject: Re: FW: tomcat 8080 thread
not reduced
Smith,
On 1/18/17 12:47 AM, smith wrote:
So the tomcat default executor will not reduce the thread count until
it reach to the max configuration?
By default, you get a thread pool that isn't as smart as an executor.
Will it reduce when it reach to max?
Not unless you use an <Executor>.
And why the default not reduce the thread?
Because it didn't do so in the past, before <Executor> was introduced.
I'm curious: if you are willing to have e.g. 200 threads available at
any time during the life of the JVM, why does it matter if those
threads are reduced during times of inactivity?
I think of threads as a resource like memory, where if you are going
to allocate X resources, you may as well allocate X resources and be
done with it. Growing and shrinking pools of things just adds
complexity and reduces performance.
Idle threads are "free" other than using a little bit of memory.
So why is it so important for those threads to stop when they don't
have any work for a while?
-chris
-----Original Message----- From: Christopher Schultz
[mailto:ch...@christopherschultz.net] Sent: Tuesday, January 17,
2017 7:18 PM To: Tomcat Users List Subject: Re: FW: tomcat 8080
thread not reduced
Smith,
On 1/16/17 8:22 PM, smith wrote:
Yes, I think thread count should be reduced when those threads are
idle
Is this right? Or it will not reduced?
Id you want Tomcat to reduce the number of idle threads, you'll need
to explicitly configure an <Executor> and use that with your
<Connector> .
-chris
-----Original Message----- From: Christopher Schultz
[mailto:ch...@christopherschultz.net] Sent: Monday, January 16,
2017 2:20 PM To: Tomcat Users List Subject: Re: FW: tomcat
8080 thread not reduced
Smith,
There are your only active <Connector>s:
On 1/14/17 1:30 AM, smith wrote:
<Connector port="8080" protocol="HTTP/1.1" maxThreads="300"
connectionTimeout="20000" redirectPort="8443" />
[snip]
<Connector port="8009" protocol="AJP/1.3" redirectPort="8443"
/>
You have not changed any settings from the default. What makes you
think that your thread count should be reduced when those threads
are idle?
-chris
--------------------------------------------------------------------
- -
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
--------------------------------------------------------------------
- -
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
-----BEGIN PGP SIGNATURE-----
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/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=dWIg
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
