Time Machine always requires deep traversal?

Pondini wrote:

All I can suggest is booting into single-user mode and deleting it via:

sudo rm -R /.fseventsd/*

The next backup will be a lengthy one (a deep traversal), but thereafter they should be quick.

That fix works great for me. I don't see anything wrong with doing it from single user mode. If there are any experts on the fseventsd daemon, would love to get an opinion.

An alternative could be disabling and then re-enabling journaling vi Disk Utility. You have to hold the option key down when selecting the file menu, then re-enabling journaling from the file menu. I'll try that and check back. But I think removing .fseventsd is the safest method.

B

Thanks for the report. Safe mode fixed the problem for me once, but I was not able to identify why.

B

bernuli wrote:

. . .

An alternative could be disabling and then re-enabling journaling vi Disk Utility.

I guess it's worth a try, but journalling and the Event Store are two very different things.

Pondini wrote:

bernuli wrote:

. . .

An alternative could be disabling and then re-enabling journaling vi Disk Utility.

I guess it's worth a try, but journalling and the Event Store are two very different things.

Oh, right. I had those 2 things cross referenced. Cycling journaling off then on did nothing to improve TM performance. Thanks for pointing that out.

I did however remove the .fseventsd directory on both my boot volume and TM usb drive. Did this from single user mode. (It is a little tricky to rm on the USB drive from single user mode.)

The big thing is to clear the the fseventsd on the boot drive.

Anyway my backup time went from 7 minutes down to 5 seconds. (11 seconds after starting up Mail, Safari and iTunes).

Good to clear once in a while, otherwise you end up with a large amount of real directories in the backup instead of hard links.

B

bernuli wrote:

. . .

I did however remove the .fseventsd directory on . . . my . . . TM usb drive.

That doesn't seem to be a problem.

The big thing is to clear the the fseventsd on the boot drive.

Yup. 🙂 I still wish there was a reliable way to fix it instead, but that is what OSX does sometimes -- such as after an abnormal shutdown or disconnection:

Jan 23 12:29:22 JPsiMac-10v8.local fseventsd[83]: event logs in /Volumes/S.Test.Backups/.fseventsd out of sync with volume. destroying old logs. (1304040 1 1645432)

Jan 23 12:29:22 JPsiMac-10v8.local fseventsd[83]: log dir: /Volumes/S.Test.Backups/.fseventsd getting new uuid: 2AA0319E-D34B-4200-BADA-2EF73ED0D7EB

Those are on an external HD, of course, but it will do the same on a boot volume if necessary.

Good to clear once in a while, otherwise you end up with a large amount of real directories in the backup instead of hard links.

I don't think it should be done as a regular thing, at least not for most users. It seems to affect only a small number of users, and as far as I can tell, only on Snow Leopard.

OK. I can't seem to get this to work. I logged into single user mode (Command+s). Prompt displayed root

I typed: sudo rm -R /.fseventsd/*

It then asked me if I wanted to delete the file (I assume becasue each was listed as rw------)

I said "y" to each file - it took ages, but eventually I got to the bottom of the list and the prompts stopped.

I reran the command, but the same thing happened. It didn't look like anything was actually deleted... I pressed cntrl+c to stop.

I then restarted as normal and ran terminal list command: It finds 13384 files in .fseventsd

What did I do wrong?

Can you please explain to me precisely what commands I shoudl type in terminal or single user mode?

Is there a way to "accept all"?

Should I have said "Y" to the prompt?

Sorry, try it this way:

sudo rm -rf /.fseventsd

What does the -rf mean?

And the * ? (I thought this would make it recursive or something..?)

There's an explanation of nearly every command right on your Mac.

Type "man" (MANual page) before the command in Terminal: man rm

As you'll see, the -f means remove without prompting for confirmation.

and since -r means remove all the contents, the /* shouldn't be necessary.

(I copied the earlier command from one of the prior posts; then just now saw the other one.)

Thanks. I just tried it, but it again didn't work.

DO I need to run the:

/sbin/fsck -fy

AND

/sbin/mount -uw /

first?

OK. That has seemed to clear it out. Same directory now just 8 items in it.

Out of inerest, when i rebooted the fans went crazt for 5-10 seconds.

Just testing with TM now....scanning...1.4 million items...preparing...637 items...backing up 540 of 663.5 MB (VERY FAST!)...now stalled...

Yes, those 2 commands are required. Sorry I should have mentioned that.

When you boot into single user mode, your hard drive comes up as read only, preventing you from modifying or deleting anything. Run those 2 commands then you will be able to clear fseventsd.

rm -rf /.fseventsd/* will then work

let me explain what that does

the rm command is used to delete a single file. by adding -rf to the command will delete a file or directory and everything below it.

Be very careful about what you type after the -rf. If you typed rm -rf / and then hit enter by accident, it will start deleting everything on your hard drive. Use control c to cancel the command at any time.

It may seem confusing, but these steps are totally worth doing IMO.

B

ZX48 wrote:

OK. That has seemed to clear it out. Same directory now just 8 items in it.

Great!

Out of inerest, when i rebooted the fans went crazt for 5-10 seconds.

Just testing with TM now....scanning...1.4 million items...preparing...637 items...backing up 540 of 663.5 MB (VERY FAST!)...now stalled...

Yes, both of those are normal. The first backup will probably be quite slow, even if small.

First backup was not so long actually:

New backups parts are much faster now, but I notice that the system still spends a long time cleaning up old backups (8 minutes). Is this because my backup HD is quite full? (95%).

Prior to removing .fseventsd, the total tiem was about 23 minutes or so.

Interestingly, the fseventsd directory already has 32 items in it.

The external also sits there for a long time afterwards churning away. Is that normal?

Wait. I've just realised TM has initiated another backup straight after the one I jsut triggered - that is why the HD is spinning. I'm not sure it is fixed. I haven't done anything in the interim and the number of copied files is now already up to 36766 (21.3 MB). :-(

Any ideas?

Straight after deleting .fseventsd all I had this in that directory was:

sudo ls -l /.fseventsd

total 8

-rw------- 1 root admin 36 5 Feb 16:23 fseventsd-uuid

Now I have:

sudo ls -l /.fseventsd

total 32

-rw------- 1 root admin 9182 6 Feb 01:42 fc007475e29523e8

-rw------- 1 root admin 36 5 Feb 16:23 fseventsd-uuid

So at 01:42 9182 files (?) were created in this directory. Is that correct?

The only things listed in the console log around 01:42 last night are:

Any suggestions would be most appreciated!

Oh, and the backup HD is still churning even though the last backup finished >10 minutes ago