Sunday, December 30, 2007
Reliability Monitor on my home PC
Though I've had Vista at work for longer than I've had it at home, the longest-lasting OS install is my home PC, a Dell XPS 710 with the factory load. Reliability & Performance Monitor on this PC has data going back to May 2nd or just about 8 months. Here are the application crash statistics collected in that time:
… and 11 other one-off crashes.
I can explain this. I was testing various software to convert videos from my camera to YouTube-suitable format and also to iPod MP4 format. One such piece of shit software called Apex Video Converter installed some ancient-ass xvid.dll that basically threw these errors every time I opened an Explorer window in a directory containing some videos. That was an annoying couple of days until I tracked this down.
As I said before, I think Firefox can be unreliable on Vista, particularly with a bunch of add-ons. But at least it saves state periodically so it's not such a big deal. Lately the problem has gotten worse, but, without going into too much incriminating detail, I have started leaving something called NOVA running which runs up Firefox's RAM usage from its usual 200MB to numbers as high as 600MB or however high it gets until I catch it. So it's easy to blame Firefox but it probably is poorly written code in the NOVA interface. Or Java. Seriously, whenever sites launch the Java tray applet, I cross my fingers against a crash.
This was the winner in the video conversion testing but mostly for convenience reasons that I won't go into and are probably the cause of the crashes. The less said, the less trouble I get myself into.
I only use IE7 at home to check my work email through OWA 2007. And, if left running for some period of time, IE won't restore and instead shows the title bar with whatever is behind it, behind it. I don't know if this happens with other websites because I browse everything but work stuff with Firefox.
I have a slight suspicion that the indexing process that Media Player uses is not stable or handles ripped audio and video material from, er, various sources without much grace. Most of the time wmplayer is doing something with its database when it crashes.
See the first (DllHost) explanation.
This is standard Acrobat Reader. A program I hate with a passion.
This I think is the HP Image Zone (or whatever) thing that installed with a photo printer I bought. I actually thought it might be nice to try another tool besides Windows Photo Gallery to manage my photos, particularly if it offered features like printing four wallet-sized photos on 4x6 glossy paper, but every time this thing indexes my pictures it gets about 96% through them in 10 minutes and then crashes. I have a lot of pictures in my \Users\me\Pictures folder but Pictures\Public Pictures are the ones from my camera. (The rest are my digital artwork and found pictures (mostly source material for art).) So something is not in whatever format HP_IZE can handle, and it crashes. Sloppy...
I had to look this one up to find that it's the McAfee Proxy Service. My feelings on the McAfee security products which came bundled with my Dell are in a previous blog entry so there's not much more to say about this piece of crap software.
Several crashes are related to the bad xvid.dll that I got from the wonderful Apex Video Converter -- they're other video conversion products that presumably called upon xvid.dll to try and convert the sample file (a downloaded episode of Grey's Anatomy for my wife to watch on her iPod, if you must know). The rest are just random crashes that you pretty much expect. I also averaged one "disruptive shutdown" a day from early August to early September, which I determined was due to problems with SATA Native Command Queuing. And, I've had five blue screens. Can't remember what I was doing and Reliability Monitor doesn't say much about it.
There you go. This concludes a fascinating examination of problems with my three computers. Stay tuned for an analysis of my event log, followed by the contents of my Start Menu. I hope I'm kidding...
Reliability Monitor on a Media Center PC
When writing my prior entry examining my work TabletPC, I thought to take a look at what Reliability Monitor shows on my home PCs. Since I spent a lot of time starting early this morning hanging out in the living room with my 7-month-old, during a brief nap I screen-capped my Media Center PC's Reliability Monitor and pasted it into a WordPad document for later analysis. (I also confirmed that there's no good way to export the data to do something less time-consuming with it.)
Since this PC for the most part spends its entire life playing TV and, shall we say, downloaded TV, for my wife and I, its key application is ehShell.exe or the Media Center 10-foot interface. This PC has had a fresh install of Vista on it since May 26th so that is the timeframe we're working with. Here is the chart of failures:
Note that this is EVERY failure I've experienced since I put my first clean install of Vista on my Media Center PC. This PC is used almost every night to watch TV and has been on a daily reboot schedule for about six months. (It will be moved back to my office within a week as I have a Linksys DMA 2100 on order -- keep an eye on the blog for details. This will move the UI work off of this PC, and make for a MUCH quieter living room.)
Why does ehRecvr.exe crash more than ehShell? What does it even do? (I assume it is the headless service that talks to the tuners, records stuff, etc.) Seeing these stats, I do remember that frequently when I resume from standby I see that error on screen, and no real descriptive information is given. Poking around on Google (constrained by site:microsoft.com; searching any process name has become a hell of shady "spyware prevention" sites), I see that there is a hotfix for those who have more than one tuner (which I do) that patches Qdvd.dll and one related to MPEG-2 decoders. Both are that wonderful type of hotfix that require a call to Microsoft. A search against site:msdn.com shows this comment on Aaron Stebner's blog (which I subscribe to and heartily recommend); following that advice I looked for ehrecvr.log but found only a zero-byte file in C:\Windows\ServiceProfiles\NetworkService\AppData\Local\Temp. I suppose these crashes will remain a mystery for now; hopefully leaving the Media Center running all the time instead of waking it up for use and letting it go back to sleep all the time, or using an extender instead of the direct UI, will reduce the problem.
Onto ehShell crashes, which are annoying because, well, you've just been booted out of the "TV" back to Windows. As per Aaron's suggestion, I found an ehshell.crash from 8:30 this morning. I happen to remember this crash because before I went to bed, I left the Media Center PC in the Video Library looking at my "TV" sharee on my primary home PC. This folder contains a lot of, uh, found TV shows encoded in a variety of ways, and occasionally my Media Center PC will crash when in this folder, attempting to index TV shows for which it may not have a codec. I resumed the PC this morning to find it in that folder, and the shell would not respond to anything. Maybe returning to a networked location that wasn't quite available yet (due perhaps to DHCP renewal delay) crapped it out? Regardless, ehshell.crash was pretty useless as all it said was "Generic failure from Win32 that did not SetLastError()" followed by a bunch of messages related to layout.
In summary, I have no idea what is causing my primary ehRecvr crash (perhaps old drivers for my tuner cards?), and I am pretty sure what causes the ehShell crashes (TV rips with esoteric codecs). Since the two crashes do not seem to occur together (according to Reliability Monitor, ehShell has not crashed since November 14th) I guess it is safe to say that, on average, my Media Center PC experiences a crash 10 times a month -- though the trend is downward lately (6 crashes per month for the last two months). Still, think about it: your TV crashing on you more than once a week, despite rebooting it EVERY DAY, is sort of like a joke you might have made 10 years ago if someone suggested that Microsoft was going to make software for TVs.
Saturday, December 29, 2007
Reliability Monitor and its analysis of my work computer
First, my usual disclaimer:
Sorry I am such a lackadaisical blogger. Bizarrely, my traffic is basically the same whether I blog or not -- between 70 and 90 visits a day according to Google Analytics; although lately the trend is downward. I will try to get a series of shorter posts out more often than every 60 days. This is not one of these shorter posts.
Is there no way to gather more usable data from the Reliability Monitor? I ran it at work and it had data going back for three months, but seemingly no export capability. To get at the data, I screen-captured and pasted each screenful of data into OneNote, used "Copy text from picture", pasted them into TextPad to fix and put back into tabbed columns, and then pasted them into Excel. Now I can pull statistics like the following:
12.0.6023.5000 (Outlook 2007, prior to SP1)
12.0.4518.1014 (OneNote 2007, prior to SP1)
various (1.8.20071.12718, the most recent, has crashed the most, and it seems that 1.8.20070.25881 was the most stable)
1.0.559.0 (2005) crashed 5 times, 2.0.6362.0 (2007) crashed once
Interwoven DeskSite (Manage32.exe)
18.104.22.168 (8.2, I think SP1)
12.0.4518.1014 (Visio 2007, prior to SP1)
12.0.6024.5000 (Excel 2007, prior to SP1)
NiAiServ.exe (Enteo NetInstall)
...and 11 other apps crashed once.
So what information can we draw from this?
It's somewhat unfair to taint Outlook by problems caused by its add-ons, or the size (~2GB) of my OST file, but 13 crashes in two months (none were logged in December, it would seem) is a couple crashes a week. And also, I'm probably like everyone else in that Outlook is far and away my most used app. So I don't know if I rebuilt my OST in the December timeframe, or some add-on that was causing instability got disabled and I never noticed, but I guess it's been better lately. I'll have to revisit after running Office 2007 SP1 for awhile.
OneNote has no add-ons, so I would have to blame its instability on a) poor handling of the online/internet/offline transition and b) LiveSessions. I have a notebook that I use constantly that's on a file share at work, a few notebooks on SharePoint that I edit throughout the day, and maybe 15% I am in meetings taking handwritten notes with my tablet. I also spend about 15 - 20% of my time in LiveSessions with one or two colleagues, taking shared notes in meetings. Handwriting is pretty stable, I think it's the LiveSessions that cause some crashing, and some of it is just random lock-ups.
Before I looked at the above statistics, I would have sworn to you that, at work, far and away my least reliable application was Firefox, and I still feel that these stats don't capture every crash or lock-up that I experience with Firefox. It is telling that Windows Reliability monitor didn't log any Firefox crash for almost two months, between 9/20 and 11/15, while I guess it was on version 1.8.20070.25881 (or maybe some interim, extremely stable version between that and 1.8.20071.2514). I like to think that I have minimal add-ons in my work Firefox but I must have enough to cause a persistent instability, like freezing when I resize the window with an IE Tab page selected, or a freeze that just happens without using Firefox for awhile.
Why do I stick with Firefox because it's so unreliable? I guess my IE 7 experience was worse, or sufficiently different and lacking in functionality as compared to Firefox that it wasn't worth it. IE 7 didn't seem to have as robust a tab saving mechanism as Firefox, and it crashed often enough that a group of tabs that I meant to leave open were gone, and that really got on my nerves after awhile.
This is a TabletPC thing that Vista is convinced has some compatibility issue. I can't remember the error but Vista seems to periodically close it and throw some error message. Mind you, this has never happened when I'm actually in tablet mode or using tablet features, it happens when I'm doing something else altogether.
I don't remember this crashing, to be honest, so I don't know why.
DeskSite likes to freeze up when you launch a document out of it -- it's not just a Vista problem, it happened a lot when I had XP, too. With the Vista error reporting feature turned on, most of the time I can actually find out why this is hanging -- typically it is Firefox but it can also be OneNote, Outlook, or Visio (which isn't really integrated with Interwoven yet and basically causes problems whenever it is launched).
I think I've already ranted about legal-specific software vendors, and I know this will be a popular search result because it contains the terms "Interwoven" and "SharePoint", so I will be gentle and note that despite my enmity towards the Interwoven product, its competitor (Open Text LiveLink eDOCS and whatever else they called it, formerly Hummingbird DM5) just a few days ago released their Office 2007 product. So, like, a full year after Office 2007 shipped. Shows you where the Open Text product is going -- down the toilet. (Disclaimer: The previous view is mine alone and does not represent the view of my employer)
If you forget about procmon and leave it running, you use up all your RAM and your computer falls to shit. Of course, you usually RUN procmon so you can figure out why something else is going to shit on your computer, so sometimes procmon is a symptom rather than a cause. Regardless, that's what's going on with my 4 procmon crashes, I think.
I don't even know what the hell this is so I looked it up to find that it's the Sprint PCS Connection Manager. I delivered (and then took back two months later) an EVDO card to my colleague and so of course I installed the software for use on the trips to and from Chicago. I have a Treo 700p with PDAnet and the PCS software, despite the 3 crashes, was like 300 times more reliable than that piece of shit PDAnet. I'm glad I can get Internet access for free and only expensed $25 or whatever to my company. But I'd way rather not have to spend 20 minutes fucking with, and ultimately rebooting, my laptop to get stupid PDAnet to work. As I have said before, PDAnet is better than not having any Internet access, but not by much.
I realize I'm detailing the apps that crashed on my work computer, which is inherently a fool's errand to begin with. But since I started it I might as well finish:
Visio: Visio 2007 Interwoven integration is non-existent, but for some reason it is enough to leave Visio.exe in memory, using 50% CPU and fucking up all other Interwoven calls, until you kill it. Hence the three crashes. I like to think that I am really good at Visio, but with 2007, if I can avoid using it, I will.
Excel: Excel is usually really good so I'm surprised to see these two crashes. I blame Interwoven for this, also. Maybe unfairly ;)
Enteo NetInstall: This company has like three support people in the U.S. and after two years of terrible service my company has ceased recommending its products. I try, on my computer, not to run NetInstall but there are some apps our IT people packaged that I need so I have to from time to time. This then leads to the inventory app running up my CPU utilization until the app is killed.
I was going to check my home PC against this, as I would think the results are very different, but I didn't realize how long this blog entry would turn out to be. I think that, in summary:
- Reliability Monitor would be even more useful if it could export its data to help you pinpoint your worst-behaved apps
- I think Reliability Monitor misses some crashes and hangs
- Firefox is totally unreliable on my work PC but is excellent at home with the same add-ons plus a dozen more; WTF?
- Interwoven DeskSite can cause unreliability
- Microsoft products crash the most :D