It's been a while since I last posted about my adventures at work with Solaris.  The optimistic amongst you may have just figured "hey, he got it working and is being snobbish about sharing the tricks to it with us."  Oh how I wish that were the case...instead, it's been dragging out, sitting there half complete while other stuff comes up.

 

It's been half complete because it was going nowhere.  And has continued to go nowhere.  But the monitor for the one Sun system has been sitting on my desk, glaring at me and serving as a constant reminder of my shame & failure.

 

High points: I got Samba working!  Single sign-on!  whoo-hoo!  Better yet, it stayed working.  Until we upgraded the DC's, which now run Windows Server 2003 R2.  Crap.

 

See, in Server 2003 R2, the schema changed to be RFC 1307 (I think that's the #...) compliant.  That means that the builtin OS tools for managing Unix identity attributes no longer work with the settings that Samba's winbindd picks up.  So effectively, Samba has stopped working as desired.

 

So after looking at the current state of things, and what was actually needed, and my available options, I decided to scrap the whole lot.  And so, in 10 minutes of using Windows to set things up, and another 3 hours on Solaris, I'm almost back to where I was with Samba for integrated authentication.

 

Steps to retrofit NIS onto Solaris:

The Intranet project at work finally shipped last Thursday before last (or something...2005-03-10).  Yay!

 

Interestingly enough, the usage stats show my load generating app with the largest amount of hits (by an order of magnitude), followed by...me.  I'll see how things go over time.

Sorry about not posting for a few weeks, but I've been really busy at work working on the Intranet stuff.  With the target deployment date approaching and the beginning of user testing against the site I've been refining a lot of things on it.  Thank goodness for VPN's!

 

On the downside, practically all my time actually spent on site during business hours is taken up by meetings.

 

This week I'm busy with a lot of midterms, and the week after that the Intranet goes live at work, so there probably wont be many posts for a bit.

I went to work again, since I'm now on my semesterly schedule.  Not much happened.  Feel free to stop reading after the title above.

Basically, I talked to one of the other IT guys, then I talked to my boss, then I talked with another IT guy, then I started reinstalling a PC (it got reassigned to a new user).  So, not much happened.

The most interesting part of the day was that we wont be using Gateway as a vendor anymore, after the shitty support we got for the NAS box (I should write a few follow-up posts to the one I wrote here about that), and now it turns out that Gateway hasn't been filling our orders correctly for the last few months.  So no more Gateway.  Yay.

The last two days at work have been pretty boring.  I just couldn't quite get into the swing of things after Finals.  Today kind of eased that transition by abruptly knocking me into whatever mode I was in during the summer.

 

I came in an hour later than usual, due to a rapid and highly planned covert operation that consisted of me whacking the alarm clock a couple of times and thus setting my alarm for an hour later when it first went off this morning.  Apparently my arm knew more about what was gonig to happen today than I did, because hindsight indicates that this was an unusually intelligent decision on the part of my (sleeping) arm.

 

Anyway, when I came in, the IT Director and Network Admin are sitting in the server room (within which my "office" resides), staring at the console.  This is A Bad Sign.  "SRVR8 went down."  The console displayed the reimaging tool off the restore cd's for that Gateway 860 NAS box.

 

It was at this point that I got knocked back into IT Pro mode instead of disaffected CS college student.

 

Apparently the box locked up between 11:48 PM and ~12:15 AM last night.  The IT Director arrives here first, and noticed that the drive mappings onto the server weren't working.  Since the system was locked, he powercycled it.  When it came back up, the OS couldn't be found and of course wouldn't boot.  And so the Network Admin was called in.

 

The Network Admin couldn't get the NAS box to boot off of what he thought were Windows Server 2003 and Recovery Tool discs, so he apparently fell back on the restore CD approach.  It was shortly after that that I showed up.

 

The good news is that the OS reinstall worked well.  The system was back online by 1 PM today.  It was also surprisingly painless, with none of the data being damaged (nice 3 hour chkdsk there...) and all the settings other than static IP, domain membership, and shares being pushed down via GP.

 

On the brighter side of things, everyone got their bonus checks today.  Now I can pay off a third of the taxes on my scholarships from last spring!

At work I've been doing validation paper work (see https://www.ntldr.net/Lists/Blog/DispForm.aspx?ID=11), but that changed this Monday.  We've had some "productivity committees" working on making the company more productive since the spring.  Apparently, the latest finding of one of these groups is that we need better access/search to old reports and data summaries that are stored on one of the servers.  Ut-oh...that means these silly (waste of time, imho, since they don't seem to be addressing the real issues) things now affect IT.

 

So I've been reprioritized to be SharePoint-Intranet-Search development guy.  I started working on the intranet site during the holiday season last year, as a low-priority, "do this when you've got nothing better to do, like reimage perfectly good PC's", type of task.  Prior to my taking over, we had a basic intranet site that didn't work, and wasn't being worked on.  So my first task was to bring it forward 4 years and sit it on modern technologies.  Which is part of the reason I chose SharePoint.  I also chose it because I hadn't worked with it before, but had seen some interesting demos, heard some stuff about it, and figured I'd treat it as a learning experience.  As it turns out, SharePoint also had just about everything we needed in a Intranet site builtin, but the IT Director didn't like the UI, so I customized that a bunch, and got it looking decent.

 

So everything was going along great with the site and the test server it was on (our production webserver was still on NT4 with IIS4 limping along), until an instrument PC died.  The instrument actually makes money for the company, so I made the call and the test pc got appropriated.  This wouldn't have been a problem, except that the backup I shot of the test system, including site config, etc. turned out to be corrupted when I went to restore it after the instrument pc problem was permanently resolved.  I know, I know, VERIFY backups...

 

In any case, shortly thereafter it came time to upgrade the servers, and then do the validation, so work on the site got pushed back a bunch.  About a month ago I started work back up on the site, basically getting it back to where I was before the whole loss of the pc.  This included getting  the few custom apps I'd written (software inventory search, computer location on map, and basic document search) back up and running.  The one issue that I ran into here was that Kerberos seemed to not be working...impersonation was always failing when connecting to the db server.

 

It turns out Kerberos is completly broken in our AD domain.  I HAVE NO IDEA HOW THIS HAPPENED.  The network admin is going through the troubleshooting guides now and should have some answer for me soon.  In the meantime, it's back to Basic auth and debugging of the search stuff.

 

The current search mechanism uses the Indexng Service builtin to Windows Server 2003, and seems to do a fair job providing search results.  Large result sets do cause perf to take a bit of a hit, but I'm hoping that those wont be the normal search types.  Profiling and focus group testing should help clarify the extent of this problem.  The biggest problem isn't going to be coding related though, it's going to be dealing with the management personel who are advocating the search, since I get the feeling that this is more a reaction to the whole search phenomena going around the web right now.  Time to go get specs locked in...(but that needs to wait until finals week is over).

Today at work I just filled out more paperwork.  It was fun!  Look at screen showing report my custom data tools generate.  Copy column 2 to column 3, put a P or F in column 4, then initial and date column 5.  Oh the fun!  But it gets better!  If I put a F in column 4, I got to put a number next to it too.  Then at the bottom I got to write the number a second time (yes!  repitition!), and write a description (copy column 1, inserting "not's" every so often).  To finish it off, I got to sign and date each page.  The magical wonder of paperwork!

 

That process was repeated for ~50 pages.

 

And more pages to come next Monday.

 

And people wonder why I've come to think the FDA and drugs are evil conspiracies.

Computer problems are always frustrating.  Especially since I seem to encounter so many of them.

 

Most recent issue: SharePoint (at work).  Last Spring Break I build a decent prototype page using SharePoint on a developmental server.  However, because of a hardware failure on an instrument, that system had to be appropriated with much haste.  The backup of the dev system turned out to bad, something that only became apparent after the original drive had been wiped.  However, it seemed like everything would be okay: the original config and content databases were still intact on our db server. 

 

Not so.  I cannot get a new install of WSS to connect to the config db.  Keeps giving "invalid account" or kills the worker process or fails with some other error (I'll remember it eventually).  The final solution is going to be to dev on another server, pull out all the documents in the old content db (using a simple app I wrote up earlier today that just pulls out the blobs), and migrate over to the production server eventually.  Oh well.  That's what the paychecks are for...

 

Update: The reason for the large time interval between initial development and now was the upgrade of all the servers to W2k3, and hundreds of pages of qualification for EVERY computer in the building (which sucked).