I seem to have upset my Linux server...

I was using tar to untar a big backup tar on a windows share, mounted in Linux, from the Linux box.

It all sort of ground to a halt!

On reboot (it failed to deal with the windows mounts on shutdown) the GUI does not load as it should but blinks until I get to a command prompt log in.

When trying the command xinit I get the errors...
---
FontTransSocketUNIXConnect: can't connect: errno = 111
failed to set default font path 'unix/:-1'
... could not open default font 'fixed'
---
Easiest fix to this seems to be to rpm the xfs rpm again... but trying this gave me...
---
installing the package requires 28Kb on the / file system
---

df shows that / is 100% used up! Is this coincidence or because of a hang or reboot in the middle of the tar operation?

So I deleted some files and there is a little space now but it still shows up as 100% and xfs will still not rpm with the same message.

So where did the disk space go? The space was 3587636 but is now 3581256. Is that at least 28 KB free now?

rc.local also seems to be unhappy and opening in text editor moans about an active swap file (red herring?).

The biggest files seem to be in root/var/spool/squid/...

A lot of new folders seem to be in there... has tar fed the untaring files through squid (or is this nonsense?).

What can I delete to make some space and hopefully get xfs and the GUI going again?

etc/termcap seems to be quite big. Can I move or delete this safely?

Luckily everything appears to still work but I need some disk space and I want my GUI back!

Please HELP! THANKS.

you can check for timestamps of the larger file and find out when exactly they were created. if the timestamp is the nearly same as the time your comp crashed you could move it to some other partition for eg. your windows partition.

So you are saying the tar 'temp' files could all be stuck in root/var/spool/squid/ ???

There are loads of folders from 00 to 0F and each of those contains a load more folders from 00 to F6. The big file is swap.state and I don't think I should delete that... I really do not want to break this box! The main folders from 03 to 0F have dates that roughly match date of crash / troubles.

How do I do a find on files created in last 3 days?

I now have 3562484 out of 3589424 (which is still 100%) but rmp xfs still says it needs 28Kb on \ !!! I have deleted a fair amount of junk already.

I will be very happy just to see the GUI again!

Thanks for your suggestions.

Regards

indeed most likely your files got untarred in root/var/spool/squid/
if you want to know how you can find files with a timestamp, i think "find" can do this, try "man find"

but first empty root/var/spool/squid/, it should fix the biggest problems. but make sure to check the timestamps, you do not want to remove the wrong files.

in the future, try to send the tar to the windows machine and use winrar to extract (if i'm correct winrar CAN unpack tar archives).
you can find winrar @ www.rarlabs.com

hope this helps

i told you to move files for that very reason. so that if accidently you move a critical file you can restore it afterwards.

THANKS!!!

In reverse order!...

Yep, winzip will do .tar and .tar.gz as well but windows does not like folders or files with names like root/, etc/, this::that, and so on. I will give winrar a go... thanks for the tip.

Rest assured I will not try this again from the Linux box!

root/var/spool/squid/ contains folders '00' to '0F', 'swap.state' file (which is big) and 'swap.state.last-clean' (which is very small). '00' to '02' are old, '03' to '0F' are on date of crash. Date stamp of swap.state is also date of crash.

What is root/var/spool/squid/ ? Can I just flush the whole thing?
What does swap.state do?

I am hoping to clear enough space to be able to rpm xfs again, hopefully letting me get back into the GUI. Once I can do that I should be able to mount the windows share again and move stuff onto the share rather than delete stuff and not be able to restore the files. I could probably mount the windows share from the command line but for now I would just like to rpm xfs and hope that fixes the GUI issue.

I don't understand as there must now be more than 28Kb of space on / now. I am guessing that it is looking at the 100% used rather than the actual free space and hence I am trying to get down to 99% used space on /.

find... yep, its something like find . -ctime +3 ... Unix / Linux command line is not my strong point! Thanks.


THANKS... No, really, honestly, THANKS!

Argh!

OK, so now I have from df...

1-k blocks 3589424 Used 3545804 Available 0 Used 100%.

Shouldn't that read Available 43620? and Used 99%?

It almost seems like I have lost permissions for the whole drive but everything else is starting up ok.

And trying to rpm the xfs rpm still says it needs 28Kb on / file system.

I don't know how I missed this but on reboot X Font Server is one of the last things to start up and it says...
---
X Font Server execcvp: permission denied
---
So it looks like messed up permissions as well?

I have had a look and permissions look ok but then I am not sure exactly what I should be looking for.

So I guess I have a few questions...

1. Why is df showing 0 Available? How do I correct this?
2. What exactly is it trying to call when it trys to start xfs?
3. How do I fix this?

I am rather perplexed now...

Thanks.

OK!

Managed to mount windows share from command line.
Moved one of the root/var/spool/squid/ folders to the windows share.
It moaned about maintaining the permissions, something about operation not permitted.
Anyway, this finally made some space on /.
rpm the xfs rpm... ok!
Tried xinit and startx and got same old error messages.
Doh!
Tried starting xfs and it sort of hung with !warning.
Started to think this was a no hoper.
Shutdown -r...
oops forgot to unmount windows share first - box hangs on shutdown.
Hit power switch on box to turn it off.
Count to ten.
Hit power switch on box to turn it back on.
I really thought it was still gonna be screwed up.
Bingo... GUI is back again.
Me... happy again!

1. I still wanna know if I can move / delete all those folders in root/var/spool/squid/ ok???

Most of them are date of crash do I figure they can go.
Two of them are 5 years old so they can go?

2. Likewise, can I delete swap.state? Will linux rebuild it?



I guess I should leave it the hell alone and be happy it works but it is still currently rather low on disk space and I guess the backup will be big until I move the rest of the folders.

THANKS for your help guys!

Cheers Tom!

G'night!