I went through a similar process with a client of ours. We did several
things to mitigate this issue.
1. Eliminate Access dbs. If you have anything running on Access get rid of
it.
2. Make sure and use the JDBC Merant 3.3 drivers for SQL - this is an update
from 6.1.
3. Check out long running db requests - maintenance tasks, utility tasks,
data imports, exports and dumps ect. Create a separate datasource for
these, restrict the number of connections and deselect the "maintain db
connections" for that db resource in cf admin.
4. Fiddle with the JVM maxperm memory, and heap sizes. Try to get them
large enough so that the heap is not being constantly resized and garbage
collection is efficient. There are a ton more JVM settings depending on the
JVM you are using.
5. Use Perfmon to figure out which resource is hanging.
6. Consider increasing the process isolation in IIS for the application - it
can tell you if IIS is having conflicts with other processes outside the ap.
7. Check the web logs for a burst of http requests at the point of failure.
One customer I know created a dynamic publishing system with constantly
updated content - then he used google adsense to serve ads and make himself
money. But every time a new page was published (new to google) the
googlebots attack to figure out which ads to serve. Since his server was
near capacity this sometimes surfaced as a denial of service type attack
(much like you are seeing) - and since many of his pages were driven by
database calls it looked like a possible db problem - or connection issue
etc.
8. Consider moving the SQL server Off of the web server.
9. Make sure and use a FQDN for the server name in the database
administrator. In other words, don't use "127.0.0.1" or any other IP
address. use "sql.mydomain.com" type syntax. In fact, put the domain in the
HOST file if you have a default win server configuration. Windows will check
the host file first to resolve the domain - which saves a dns request. JDBC
will attempt to resolve an ip into a domain name using reverse lookup before
making a connection. In many cases this results in a lengthy initial
connection request. Having a guaranteed resolution to a flqn (a la the host
file) eliminates this problem.
That's my short list - anyone else?
Mark A. Kruger, CFG, MSCE
www.cfwebtools.com
www.necfug.com
http://blog.mxconsulting.com
-----Original Message-----
From: Bill Grover [mailto:[EMAIL PROTECTED]
Sent: Tuesday, August 31, 2004 8:11 AM
To: CF-Talk
Subject: RE: 100% CPU
Ian,
We have seen this behavior as well. When we start experiencing the page
slowdowns and notice the jrun CPU very high we will stop and restart the CF
Application Server, this "fixes" the problem. Yes the pages are down for a
few seconds but in our environment this is not as big a problem as having
the slow page response, or the page never coming up at all.
______________________________________________________
<file:///E:/EUColor.gif>
Bill Grover
Manager, Information Systems Phone: 301.424.3300 x3324
EU Services, Inc. FAX: 301.424.3696
649 North Horners Lane E-Mail: <mailto:[EMAIL PROTECTED]>
[EMAIL PROTECTED]
Rockville, MD 20850-1299 WWW: <http://www.euservices.com/>
http://www.euservices.com
______________________________________________________
-----Original Message-----
From: Ian Buzer [mailto:[EMAIL PROTECTED]
Sent: Tuesday, August 31, 2004 5:00 AM
To: CF-Talk
Subject: 100% CPU
Hi all,
We're running a CFMX 6.1 / SQL server 2000 app on a 1.6GHz AMD server with
1.5Gb of RAM. The app gets about 2.5 million page requests a month,
peaking
at about 7000 page requests per hour.
Since yesterday morning it's suddenly started to misbehave, with the
processor regularly maxing out at 100%. It will run clean an fast for 20
minutes or so, then suddenly requests will start to queue up and the
whole thing grinds to a halt. Usually it recovers, but sometimes the
service needs restarting to get it back.
The app is well behaved, with most requests taking about half a second.
There seem to be no particular long running requests that trigger it.
This happened a couple of months ago and I restricted the amount of memory
SQL server used and increased the RAM on the box to 1.5Gb. It currently
appears to have about 3-400Mb free.
I've installed the CFMX updater (released a few days ago), reinstalled
ColdFusion, scandisked the drive and defragged.
I wonder if anyone can give me any clues on what else to look for. Surely
we
should be able to get more traffic out of this spec machine?
Many thanks
Ian Buzer
_____
[Todays Threads] [This Message] [Subscription] [Fast Unsubscribe] [User Settings] [Donations and Support]
