« March 2008 | Main | May 2008 »
April 30, 2008
Miata Project Update
I just had to take a few minutes and tell you about my latest experience with my new (to me) Miata convertible. It has just under 90,000 miles on it, so was in need of a timing belt replacement. In a crazy moment I decided to do this myself. So, parts were purchased, a manual was acquired, tools were organized, and I tore that thing apart!
I replaced the timing belt, crank seal, tensioner pulley, idler pulley, water pump, radiator hoses, drive belts, and who knows what else. Put the whole works back together according to all the information, and that darn thing would not start! Tore it apart again, reset the timing belt, same thing - no spark, no firing, nothing.
Forturnately, my neighbor's son is a mechanic - an excellent mechanic I might add. He agreed to come over and look at my handiwork. Well, he discovered that a sensor had been knocked out of place. He reset that sensor to specs and the Miata started right up, purring like a kitten!
Today, I'm going for a ride even if the weather is rotten! You know, I learned a few things...
One: don't do a large project like that on a car you just bought - go out and have fun with it
Two: you learn a lot about a car by replacing the timing belt
Three: I'm not as good a mechanic as I thought!
Honk if you see me and my little Miata out and about!
Posted by rossm at 10:14 AM | Comments (0)
What's Happening With Astra?
Astra schedule has taken off like crazy again. The events folks are working overtime to get all the pieces put together to allow all events from every corner of campus to be scheduled through Astra. This is not a small task! For my part I've been installing the Astra client on various computers all across campus, adding events users, and adjusting user privileges. Just for fun, one of our services - the one that moves Banner data into Astra - decided to quit working. It really fooled me because the system showed that is was working normally. However, repeated reminders from the users told me it really wasn't working. By stopping and restarting the service, everything got back to normal. So, the fix was easy, the diagnosis was hard.
Next on my list of Astra tasks is to get the web email notifications working. That should be finished by tomorrow. Once this is working, it opens a whole new can of configuration worms, as many rooms on campus will need to be set up to notify the proper individuals when there is a request for an event.
All in all, Astra is really turning out to be a very positive addition to campus. It truly is a HUGE advantage to be able to look in one place and see everything that is going on around campus. Also, the scheduling engine that does the actual scheduling of classrooms is a great advantage.
In the near future, I will be adding in automatic notifications to persons such as custodial, HVAC, and campus safety when events requiring their services are scheduled. This will help to give these folks better information on what is required all across campus.
Posted by rossm at 9:58 AM | Comments (0)
AERO Gets A Friend
First of all, let me update you on AERO. It has survived it's upgrade to 10gR2, and we've had some time to shake the bugs out of it now. Enterprise Manger (EM) is running, and Dale and I are learning how to take advantage of the many features of EM. Let me give you one example. When AERO was first started up I set the baseline on SQL Response - it was right at 16 seconds. I'm not exactly sure what this means, because the average wait time for SQL queries certainly was not 16 seconds, but anyway that's the baseline. Using a feature called ADDM, I was able to identify about 6 of our worst SQL tuning offenders and fix them right from EM - usually by adding an index. The baseline is now less than 2 seconds after performing these tuning tasks. Also, we now have RMAN backups going and Thursday night we will have multiplexed redo logs. This database is as stable and well tuned as any we've had on campus, thanks in large part to the new EM.
OK, so AERO is doing well. We have a sibling db running on a separate server at version 10gR1. This db just happens to have the correct character set for the Viking application - what a stroke of luck that was. Michael Ellis and I have been testing the migration of Viking to this 10g platform. Right now we're working on creating a DAD which will use AERO's web server to connect Viking users to the 10gR1 instance. It should be doable, but is of course a challenge. Yesterday I learned about tnsping, which can be run from a Solaris command line and can be used to test the connect string for the DAD. The one we had wasn't working, so we are on the right track to get that going. The actual migration of Viking might require the expertise of the Viking techs because it still uses long raw data types, which don't make the export/import jump to hyperspace. As luck would have it, the 10gR1 EM is totally hosed, so that's another project to keep me off the streets!
Posted by rossm at 9:40 AM | Comments (0)
April 16, 2008
Let's Go! Summer Was Here
There was a brief sighting of summer - last Saturday. I did not let the opportunity pass me by. Here is a picture of my new mode of transportation - a 2002 Mazda Miata (taken on a cell phone from my friends' BMW 330).
OK, so if this works you can see my newest toy, and I learned how to add an image to a blog. It's not a bad picture because you can't see me too good.
Posted by rossm at 9:31 AM | Comments (0)
10gR2 - We Love EM!
Yes, we love Enterprise Manager! Dale and I have continued our quest to whip our production database into shape. This week we did a lot of minor tweaking such as adding indexes based upon recommendations from the ADDM (whatever that stands for) section. It will identify sql queries that use up a lot of cpu or database time and offer suggestions for improving the execution. In three separate cases a new index was recommended, and in one of these cases adding the index reduced the execution cost from over 5,000,000 to about 8,100 (this index saved multiple full table scans, which is a costly operation on a big table). You can create the index right from EM.
Another important improvement completed this week was activation of the flashback area. The neatest part of this is that we can now recover a table that was accidentally dropped in about a minute. Without flashback, this takes about a week! The bigger, but not as obvious, advantage is that we can flash back the entire database for about 48 hours if something major happens. In other words, the whole database can be recovered without resorting to the archive log files, at least for a short time.
The last piece I'll write about here is setting up a good backup. This requires the flashback area to be in place. I put a job together that ran last night, but it failed because I'm not too good at math, I guess - tried to put 40 plus GB of files into a 5 GB space. Guess what, it doesn't fit. So, about 10% of the backup got done before it totaly filled up the space. Anyway, the backup will work once we get the appropriate amount of disk space set aside for it. Clayton is working on that for me.
Using Enterprise Manager in the 10gR2 environment is really helpful. EM tells you about many, many items in the database that are really hard to see using a command line. Some examples: identifying tablespaces that are running out of disk space, tablespaces that need to be shrunk, take tablespaces off line or put them back online, sql statements that need tweaking, important alert log entries, and on and on.
Posted by rossm at 9:12 AM | Comments (0)
Astra Is Moving Along!
The Astra implementation got a huge shot in the arm this week. We had a meeting of the Astra administrators and were able to decide to move all event scheduling to Astra. This means a lot of work, but it also means we'll finally be looking at one place only to see what spaces are available all across campus. This is neat. Very soon we'll be releasing the web piece which will allow anyone to request events and see what rooms are available from the web - Very Cool!
Once we get all the events pieces put together, then the campus will really be able to take advantage of the application, and it should cut down on double bookings and a lot of confusion. I am constantly amazed at the number of events that are scheduled across campus - it is a lot!
(By the way I attended one event last night at the WUC Pacific Room - Step Afrika - it was really good!)
Another piece that will be very useful is the notification of service persons such as Custodial, Setup/Takedown, Public Safety and Heating/Air Conditioning. Also the combined information will make it easier for these folks to track events directly.
It's good to have this Astra train moving down the track again!
Posted by rossm at 9:00 AM | Comments (0)
April 9, 2008
Astra Events - Here We Come!
Well, Astra configuration has turned another corner. Jill has begun entering events for the WUC in, and there are a lot of events visible on the Astraweb. One problem is still with Astra to fix, and that is the "collection" issue I wrote about previously. Basically, the event confirmation doesn't bill correctly if a collection (a group of resources) is reserved. Aside from that, Jill has the configuration done and can now schedule events with Astra. The next step is to meet with the key Astra users to determine how to configure the web event request notifications. In other words, how do we complete the event scheduling process once a request is received? We need to make sure that the building managers, event schedulers, and service providers all get the information they need. We will be meeting early next week to figure this next phase out, and get Astra properly configured. After that is completed we'll be moving on to include all other event scheduling areas (Hamersly, Athletics, Music...).
The implementation of Astra has proven to be a monumental task. It is really amazing to see what all is going on behind the scenes to schedule events for the campus community! Wow!
Stary tuned as we bring the Events part of Astra online.
Posted by rossm at 9:57 AM | Comments (0)
10gR2 Gets Annoying, Possibly Fixed
So often, upgrades drive you crazy while you chase down annoying problems that crop up. You really appreciate the new features, but it sure comes at a cost. Unfortunately, this upgrade has been no exception. Right after the upgrade, the Enterprise Manager wouldn't start up. Turns out the installer puts different time zone files in for the EM and the Java virtual machine. Must bounce the db to fix, which we did when Sundown was down. Next annoyance: approximately every two days or so the Listener hangs. At first, it appeared as though the database itself had expired, but no, the listener just hung. Bill figured out that somehow two listener processes had gotten started - cause unknown - but, killing those processes and restarting the listener put you right back in business. Of course, this got old the third or fourth time it happened. This morning, I put in an SR to Oracle and they got right back to me suggesting I put this command into the listener config file: SUBSCRIBE_FOR_NODE_DOWN_EVENT_LISTENER=OFF. Well, we had tried this before and the listener wouldn't start, but I had received better directions on where to place this command in the config file. Sure enough, Bill stopped and re-started the listener and it started right back up with this line in the config file. So, hopefully, this annoyance will also be behind us.
I've got one more thing to fix right away, and that is the flashback area. It had been set to start up with the last restart of the database, but it didn't. This requires a bounce of the database to start, and will have to be done soon. The flashback area is very similar to the Windows Recycle Bin. If an object is mistakenly dropped from the database you can just go in there and put it back. Very slick, but of course it doesn't work unless configured.
So, aside from me complaining about these annoyances, the 10gR2 version of the database is a real step forward. Once these few items shake themselves out, then I'll get busy doing some fine tuning and tweaking to really make it hum. That will be a lot more interesting!
Posted by rossm at 9:37 AM | Comments (0)
April 2, 2008
10gR2 Has a Wild Week
Well, our server that never crashes did just that late last week - last day of Spring break, good timing. Of course that brought the database, listener, the whole ball of wax, down. But, while the database was down I copied over those two time zone files I got from Oracle Metalink support and by golly, Enterprise Manager is now functional. What a relief that is! At least now I know what all needs attention. Dale and I moved a bunch of objects out of the System tablespace. Somehow, we got a number of users with System as their default tablespace, which make absolutely no sense! Anyway, we just altered those users and moved out all of the offending objects.
We got to learn the syntax for moving an Index to another tablespace, which is:
alter index [index name] rebuild tablespace [new tablespace name].
The syntax for moving a table is:
alter table [table name] move tablespace [tablespace name].
On Thursday night, I'll be doing another 9:00 pm to whenever shift to make some modifications to the 10gR2 production instance. When I started it up during the upgrade, a wrong version of the initaero.ora file was used, which resulted in our control files and archive files staying in their old location. One of the main things I want to do this month is to move those over to the proper location. Also, we have maxed out our number of data files at 200, so that needs to be increased. I want to increase the processes parameter to 500, too. When we lose connection with OUS databases, our processes spiral upward and it locks up the database. Hopefully, this will help with that problem.
While I have the database down, I also want to do some maintenance on the Apache webserver logs, and take care of an ons.log problem from the upgrade.
It will be a full night of maintenance, but not nearly as demanding or stressful as the upgrade, I hope!
So far, the 10gR2 database has proven to be a nice, solid Oracle database for our production environment. The Enterprise Manager upgrade was worth it alone, it is really slick. In addition, I have installed and implemented Oracle's "Software Configuration Manager" which is a part of the Metalink support. It sends information about the database to Metalink, and Oracle makes recommendations about items that need attention on the database. It is actually very helpful.
Much going on with the database still, and plenty to keep a guy like me happily busy for some time. I am anxious to get these fixes in place Thursday night, it should be fun.
Posted by rossm at 9:47 AM | Comments (0)