server woes: when to throw in the towel

Sorry, all I heard was "I'm trying to make my life miserable"

Too bad these companies basically killed themselves with 'big business' support policies that don't even fit with the modern world in fortune1000 companies, let alone the little guys.

 

Nothing to do with Sarb-Ox, just looking for better fault tolerance. We may be able to split our environment over 2 locations, we'll see where we end up.

Regular MS support for Exchange can be tricky. They've outsourced a lot of it, it used to be that the Exchange Enterprise support had access to the "real" Exchange guys (developers, etc.), but that doesn't seem to be the case since the switch. We have a complex, multi-forest environment. After the 2nd call to support where we literally had to spend a day explaining everything to the 1st level drone, we popped for Alliance support. Now, we have a dedicated 6 person team that knows our environment and we call or email direct. The guys we deal with really know their stuff. We are also fortunate to have a good Exchange (2003) environment, we were able to build it from scratch the right way, including Active Directory.

We're migrating off of HP OpenMail. 6000 home office users are done, we should have all 14,000 in the field offices done by the end of the year.

I find that the MCSE tests are a little more difficult than the NT 4.0 days, but not too bad, as long as you are working with the product every day.

So who's going for their 2003 MCSE now? Just me and Enzo? Anyone want to start a pool on who finishes first? I need some motivation.

 

Hey you are in WI. My in-laws are in Stevens Point (I think that is close to you). The good ol MCSE tests I hate them. I have committed to going through the boot camp just to get it done with.

 

Close enough. Stevens Point is in the middle of WI, I'm in the SE corner.

I hate them too, but what's a consultant to do?

 

update: Friday - spent lots of time replacing parts as noted prior. Yesterday was my Monday, after having Monday off - sorta. I worked, but the office was closed. Tuesday I worked with the installers to rework Oracle 10g after wiping and recreating the array in the storage unit. We spent ~10 or 11 hours on it yesterday. Lots of things needed to be undone manually and it took a bit of time. Today, works all fine, wait... doing a tty report and there's the same problems again. So... coming up on 6 weeks now (Friday), they are coming out again this evening with more people to observe what the heck is going on.

When will they throw in the towel and give us a new one? Or when will we throw in the towel and go somewhere else?

On a good note: FarmerDave's office got a new server in and it's working just fine fine fine - at least according to his text. =)

 

Yup. Everything went as planned except that

1. The monitor we were using at the server was so old, that the video card on the server didn't support a low enough resolution... so we had to buy another monitor (luckily this happened right before lunch so I got to take a break and meet the DFW crew for lunch.)
2. The CDROM drive doesn't work, so we had to install apps by copying installation files from CDs on different computers, transferring data to the server HD, and installing that way.

Could have been much, much worse.

 

Yup, but you are thinking like a techie, making things work. Good to hear.

A rep from the company is here now, not the repair tech either, an internal tech. More of them will be out tomorrow. This is interesting.

 

I finished last year.

 

I'm in Manglement now, I'm required to let my tech skills deteriorate

All my certs have expired...

 

You should at least keep current. That way you know if your admin group is dragging their feet or in over their heads. Plus, should you move, it makes you more marketable.

I no longer do Cisco stuff but I have kept my CCNP current because you never know. I was at the IBM sterling forest DR site in March and I had to reflash the routers and add static routes on the switches . Had I not kept my NP current, I would have had to make a call of shame to the LAN/WAN group.

 

well, the scsi channel array is going away. we're sending it back soon. we're getting dough back as well, so just under 12K to spend on credit. now... I need to get a fiber channel array instead.

the base is two PE2850's* which have PERC4/DC's in them. Those will go away. Now... what do I need?

* yes, I gave away the OEM to those that know. I've had it with the PV220s that has been putting up fits.

suggestions? looks like this time I'll be setting up the Oracle 10g on the cluster using ASM. (not OCFS) One other question: this was set up as RAW. Do you feel it is a huge advantage having it setup that way? I do, but I'm being pushed to make it "browsable" by the boss. He wants the freedom to drag and drop files in there. (dangerous, but give me some reasons why!) I understand that RAW doesn't have file system checks like ext2 or even a hefty permission level like ntfs, but it will depend on how much latency we're talking about here on the file system itself.

 

If he is wanting to browse and drag files, You should go raid 1 with the OS, and raid 5 for the DB and raid 10 for the redo logs if you have the money to waste on disks. Is this connected to a SAN? I am running Oracle on the inside parts of the IBM fastT drives

 

That's a DELL OEM controller IIRC. Sorry I am late to the thread.

 

I don't want him to drag and drop files on the storage server. That's pointless, especially in a redhat evironment. It's trying to convince the guy that RAW is great, but... not for browsing due to the lack of file checks. It is quite fast.

The configuration was this:
2 - dual 3ghz/4GB front end boxes running mirrored 36GB drives on the internal PERC.
1 - external storage array connected via SCSI cables (one from each node of the cluster), using 7 drives in RAID 5 and one hot spare.

 

this is the first mention of the OEM. I was keeping it quiet. =)

 

LLED is not a good choice to run Oracle. The drives are not up to muster unless you are running ultra 320's 15k, HPUNIX or IBM rs6000's are better suited for it.

 

Are you putting the redo logs on a the raid 5 too? You will see performance issues, even if they are on different LUNS.

 

yup, the 220s storage array was (is trying to) run U320 15K's. great and fast, but... now, after 6 weeks they are wanting me to try something else... setting it to U160. we threw in the towel and are trying to go with something else.

the fc drives that are out there - EMC arrays, they are all 7200 SATA drives and that's it? seems like they would take a beating with oracle.

 

haven't gotten there yet. we have gotten as far as installing 10g, setting up a small database and doing an import... and voila... system data corruption errors and the tty reports have endless scsibus reset errors, drives start to fail, go offline and the database is not just offline, it is gone from existence.

I might have to hit you up for a call later to see what you might suggest. I have a feeling we'd need to stick with the same OEM though. IBM would be great, but... we only spend just shy of 40K for this setup with oracle licenses.

 

SATA's usually have pretty good I/O wait times but a Fiber attached SAN really is the only way to run oracle without performance issues. Try running the OS raid 1, DB raid 5 and redo logs on raid 10 using different LUNS on different attached storage units. This will still allow you to run the LLED server and not take additional IT budget hits.

 

Call me anytime. I will PM my cell.

Check my subsequesnt post about hardware and raid setup's.

Oracle does not like imports. It likes restores.

 

interesting. but the import is just data into a blank database - migration from 8i (sun solaris) to 10g (rhe3).

danke schoen. will give you a ring in a while. gotta catch up on a few calls first.

 

Hey, Randy, guess what? All the scanned documents we had are still there, but the guy didn't name the server with the same name and folder configuration as the old one, so now, instead of using ACT to search by client name and clicking the attatchment in their profile, we get the pleasure of changing the shortcut string so that the attatchment can be clicked and viewed. No way I, or the guy I hired, can figure out to change them all at once. Isn't that special?

 

while I cannot help with ACT, I can give you this information. the server's name was called "server" and the share was called "office". note: that's the share name, not the folder name. here's the info from the scan done while I was there. =)
Image Unavailable, Please Login

 

Yeah, I know, I already knew exactly the names of the server and it's relevant folders, but the guy is telling me it would screw up alot of other stuff if we went back and renamed the new server the same as the old one.

I fail to see what the problem would be but at whatever rate he is charging me, and my new perspective on his abilities, I don't feel like paying him to explain it to me.