Disk memory black hole! Something hidden writes to HD (Snow Leopard)

[sybawi]

Update: we're back in business. Deleting iStatMenus and all its files did the trick to free swap.

Not to say that that program was the cause, too. Perhaps it was the interplay of low RAM and some other program's swap issue that held all in a loop when the computer went to sleep.

Quite a relief must say to not see that free storage decrease rapidly like a timebomb counter, and to be able to just drag a window around in a normal way. I ran- sorry opened, installed and then ran OnyX, first while still in safe mode, then, after success in SMART check and having Automation run (=maintenance, rebuilding & cleanup), ran it all again in normal mode. All successful. And it didn't even take that long.

Activity monitor is top on the process list again (as you can read, before it was SystemUIserver, hyperactive apparently); Sys.Memory: Page ins: 659.3, Page outs 0. Swap used: 0.
'Black hole' is gone, captain.

Thanks guys for the help!

Kind of ironic that the program I used to start attack solving system problems became the problem itself. Like accidentally getting a thermometer stuck inside the patient's throat

Still, i suspect my harddrive will be full of bad sectors/blocks since it's old (4+ years) and prior to this swap leak there have been a lot of crashes ('black curtain' ones, which often is RAM failure I believe) and rough restarts following, which can never have been good for writing & rewriting.. I'll have a look at lifesabeach's recommendations for checking NFTS. But perhaps prior to updating this system (10.6.8) I should get me a new HD as well..

 

[MacInWin]

Onyx ran a check on the drive when you started it up, so if it didn't find any errors, your drive should be ok.

I use Bjango's Istat as well, and I have no problems. It sounds like your install went south, leaving you in a weird state, partially installed, partially functioning. So when you rebooted to wake it up, iStat tried to load, couldn't, so it kept trying (using memory) until it triggered swap, until the drive filled to the point where the whole system slowed down.

Re-download iStat and reinstall it completely. It should work just fine. Make sure you get the right version for your OS level if they have different offerings.

 

[Lifeisabeach]

Onyx ran a check on the drive when you started it up, so if it didn't find any errors, your drive should be ok.

I have to disagree with this. Onyx relies on the short SMART testing that reports using the manufacturer's guidelines. Everything I've read about SMART testing is that if you fail a SMART test using the "short" test, your drive is in critical condition and about to fail at any moment. Doing a "long" test using more advanced testing tools will give a clearer picture of a drive's health. You can easily pass a short test, but your drive still not "quite" on the edge of death to trigger a fail yet already exhibiting issues. Even the long test isn't foolproof. I had a drive crash on me last year and it even passed the long test while it was doing nothing but thrash and grind.

 

[MacInWin]

Onyx also does a Verify on the disk, exactly the same verify as you get with Disk Utility if you choose Verify Disk. You can see the results of that test under the "Verifying" tab and then "Structure of Volume." Basically Onyx invokes fsck_hfs to test the drive.

Now, nothing will predict when a drive will fail. I once ran a drive check only to have the drive heads crash completely, destroying the drive, less than 5 minutes after the drive test reported back that all was well, healthy drive. So I now make two separate backups onto two separate external drives daily.

 

[sybawi]

Re-download iStat and reinstall it completely. It should work just fine.

Did, it's running fine now. (I didn't even have to re-enter my serial, which made me wonder if I indeed did delete all files or if Bjango kept a 'cookie' on me. But that doesn't matter now.)

Even the long test isn't foolproof.

Now, nothing will predict when a drive will fail. I once ran a drive check only to have the drive heads crash completely, destroying the drive, less than 5 minutes after the drive test reported back that all was well, healthy drive. So I now make two separate backups onto two separate external drives daily.

Guys, you're scaring me. I should get a new HD soon, I get it. Or two. Three? Backup, archive, redundancy.. Hard disks break, and fear sells.

One last question: what's best practice for keeping sectors & blocks clean, organized and not too fragmented on HD's? Regular disc repairs? Restarting every now and then? Moving out a user's folder and copying it back in?

I guess that will be another thread

 

[MacInWin]

Great job on the reinstall. There probably was some .plist or config file that was kept from the partial install.

As for HDs, they fail. Period. There are two types of owners of drives: those who have had them fail and those who will have them fail. The groups overlap.

Backups are a good thing. If you aren't making them, start now. Get an external drive, usually one twice a big as the internal, connect it up and use TimeMachine as a start. You can get Time Machine Editor if you don't need TM to make hourly backups. I use TME to set TM to backup my system at noon and midnight. I don't need it more frequently than that.

Now, I'm one of those folks who have had a backup drive fail, so I make two backups. I use Carbon Copy Cloner to make a fully bootable backup to a second external drive. That way, if my internal drive dies, I can boot from the external backup and be running right away while I get a replacement drive in to replace the dead one. Once the replacement is installed, I can then use CCC to clone the external back to the new internal and I'm back where I was.

So, backups are good, but you don't have to be as paranoid as I am to at least protect yourself from drive failures. ;D

 

[sybawi]

Thanks on the HD backup advices, I indeed do run backups with Timemachine regularly already. Although I don't trust them too much either. I might be mistaken, but I tend to think it also backs up the hidden problems. So sometimes I just copy the files I want to keep and back that up (perhaps better: archive).

It's high time I should clean up this old buddy before/ while I update it to a newer OS. And clean out unnecessary leftovers from any old write. Hence my former last question on maintenance, not back ups. What's best practice?

 

[MacInWin]

Just run Onyx about once or twice a year. That should do all you need to do. Keep the HD below 85% full will help performance. Reboot every few weeks to clean up cache. That's about it.

 

[Lifeisabeach]

Onyx also does a Verify on the disk, exactly the same verify as you get with Disk Utility if you choose Verify Disk. You can see the results of that test under the "Verifying" tab and then "Structure of Volume." Basically Onyx invokes fsck_hfs to test the drive.

Yeah, but even that's not foolproof. That "just" checks the integrity of the file system, not for failing physical sectors, right? Although... I do suppose if a drive is "slowly" failing, then that would be evident when "verify disk" fails often enough.

Now, nothing will predict when a drive will fail. I once ran a drive check only to have the drive heads crash completely, destroying the drive, less than 5 minutes after the drive test reported back that all was well, healthy drive. So I now make two separate backups onto two separate external drives daily.

Yeah, more or less the same here, though my secondary backups aren't done quite that often. My Time Capsule finally bit the dust last week (drive is probably ok, need to pull it still), and I was crossing my fingers hoping the internal drive wasn't going to follow before the secondary was updated.

 

[MacInWin]

It's hard to tell what fsck_hfs is doing when invoked by Onyx. Here is the Apple man page on it. Depending on the flags used by Onyx when it invokes it. Note the -S flag:

-S Cause fsck_hfs to scan the entire device looking for I/O errors. It will attempt to map the blocks with errors to names, similar to the -B option.