Linux question of the day

paul78

Yeah - I didn't know the answer either so I didn't want to totally give it away -

But I will admit that I cheated and looked at the bash source code and saw it as an available redirect device.

ChooseLife

paul78 wrote: »

Oh - how cool - I didn't know bash had a builtin tcp device file. Good one.

*thumbs up* I like your subtle hint
Glad you enjoyed the challenge

ChooseLife

onesaint wrote: »

it's not apparent unless you've read about it someplace

Yes, it is true for this question, although comments show that some fellow TEers took on the challenge and did "original research" (citation needed).

ChooseLife

onesaint wrote: »

Great reference thread and question!

hiddenknight821 wrote: »

Great thought-provoking question again, @ChooseLife.

Thank you, fellas. Your feedback motivates me to make the effort to come up with interesting questions...

UnixGuy

I had to search for answers for the last few questions, good job guys

New question...

Q) We have a script called "backup.sh", we need to schedule it to run at 10:00 pm every 4th Saturday of every month. How can we do that?

log32

Three approaches to achieve this in my opinion.
1) in crontab, 0 22 24-31 * 6 backup.sh
2) edit the script to check for the current day and if it matches the last Saturday of the month only then it gets executed.
3) create a crontab with the && operator, the first script checks for hte last day of Saturday, if it succeeds, then the other script gets executed. e.g. :
00 22 24-31 * 6 check_for_saturday.sh && backup.sh

UnixGuy

@log32:

Now this is the trick. This is a common mistake:

00 22 24-31 * 6 <== this will execute every Saturday (because day of the week = 6) AND everyday in the range 24-31 ! So this will execute whenever any criteria is met, rather than executing in a mutual exclusive manner.

So this will execute every Saturday and every Sunday, Monday, Tuesday , Wednesday,..etc in the range 24-31

ChooseLife

UnixGuy, that's a good one. As I understand, log32's solutions #2 and #3 are the right answer, but if you know a better trick, please do share!

onesaint

I think I would do a combination of cron and scripting. Maybe run the cron job daily and do a date check to see if the current date is a Saturday (date --date="Saturday") then check if the current date is within the last 7 days of the month (with cal maybe?). If it is, execute the rest of the script, if not, exit.

UnixGuy

@ChooseLife:

They will work but not always. From 22-31, what if there are two Saturdays? the script will execute two times


We have to test if this is the fourth week(i.e. the last seven days of the month) and it is Saturday. We need to keep in mind that some months are 30 days, some are 28 and some are 31.


The test can be either in cron itself or in the script. I prefer to put the test in the script itself. For me I'd run the script from 22-31 and I'd test if the day is Saturday to run it. Maybe create a counter somewhere in a text file to check if the script has run once before or not to avoid two saturdays situation.

[COLOR=#000000][FONT=verdana] [/FONT][/COLOR]00 22 22-31 * * [COLOR=#000000][FONT=verdana]test 'date +\%a' != Sat || /backup.sh[/FONT][/COLOR]

The above code will work, but it will execute two times if there are two Saturdays in the period 22-31. We have to check for this in the script itself. One way to do it is to have an empty text file as flag with the value 0. if it's changed to 1 then the script has run.

Here some discussions:

How to Set up a cronjob On 4th Sunday of every Month? - Page 2 - The UNIX and Linux Forums
Run a cronjob task every 3rd Sunday? - The macosxhints Forums
Aix - View topic - cron for 4th sunday every month?

ChooseLife

UnixGuy, I should have made it clearer that I was referring to solutions #2 and #3 above more holistically, without looking at exact numbers there. What I meant was that checking the date would have to happen as part of the command sequence executed by cron (either within the invoked script or outside of it) and not as part of cron time parameters (first 5 fields)

UnixGuy

@onesaint & @ChooseLife: true both are correct answers

log32

Awesome questions & Answers, keep them coming!

ChooseLife

ChooseLife wrote: »

Question: How can you download a file from an HTTP server using bash without netcat/curl/wget/w3m/lynx?

As a follow-up, here's a somewhat related article:
Netcat without -e? No Problem!
which I hope some of you may enjoy reading

ChooseLife

Here's another one. This question, in my opinion, can differentiate between real experience and book knowledge. An experienced sysadmin who has encountered this issue at least once before should have no problem figuring it out. Books, on the other hand, do not teach you this - if they do, surprise me!

Unfortunately, this scenario is not reproducable locally unless I provide instructions that give out the root cause. So ask me questions or ask me to type commands on this pretend system.


Question: Suppose you are on a production server and are trying to create a new log file. However, you encounter the following behaviour:

$ echo Test > /var/log/my_new_custom.log
bash: echo: write error: No space left on device
$ su
# echo Test > /var/log/my_new_custom.log
bash: echo: write error: No space left on device

Your little investigation reveals the following information:

$ ls -al /var/log
drwxr-xr-x 18 root              root         4096 May 17 10:24 .
drwxr-xr-x 12 root              root         4096 May 17 09:40 ..

$ ls -l /var
drwxr-xr-x 18 root root     4096 May 17 10:24 log
drwxr-xr-x 10 root root     4096 Mar 17 10:22 spool
drwxrwxrwt  2 root root     4096 May 17 09:54 tmp

$ df /var
/dev/sdb1                156185596 42384420 113801176  28% /var

Why is the system complaining about lack of space? What command can you execute to find the root cause of the issue?

I have a tiny hint for this one, should anyone need it. Another hint is embedded into the question.

UnixGuy

###

Interesting problem, I run into it from time to time with servers that run Oracle databases.

ChooseLife

UnixGuy wrote: »

Maybe...

Of course you know the answer, UnixGuy, I wouldn't doubt it

paul78

Good one. I pm'd my guess so I don't hog the answer attempts.

ChooseLife

UnixGuy wrote: »

Interesting problem, I run into it from time to time with servers that run Oracle databases.

I first ran into it with a misconfigured mail server and a bunch of other systems since then

ChooseLife

paul78 wrote: »

Good one. I pm'd my guess so I don't hog the answer attempts.

And you've got it right!

onesaint

When I was first starting to learn Linux, the guy teaching stressed the point of using mv when i couls as opposed to cp because you'll use up the items in question (if I'm right).

ChooseLife

onesaint wrote: »

When I was first starting to learn Linux, the guy teaching stressed the point of using mv when i couls as opposed to cp because you'll use up the items in question (if I'm right).

Looks like the question was not as tricky as I hoped... Fine, let's consider this one solved

paul78 or onesaint (or anyone else), care to answer this part to finish it up:

What command can you execute to find the root cause of the issue?

(UnixGuy did in his initial response)

paul78

Hmmm... Not sure I know of top of head but I would probably start with fsck. And then used find -size to find all the small files. Usually there's a bunch of small temp files lying around if its a email or spooler.

onesaint

# df -i
would give you node info (I just used it the other day when a server went RO for no reason!).


I've got (maybe an easy) one for you all:

Let's say I accidentally did an rm -rf on ../httpd/access_log but I still need the file.

Is it recoverable? If so, how?

ChooseLife

Ok, I'll take these two combined as a full answer

onesaint wrote: »

# df -i
would give you node info

paul78 wrote: »

And then used find -size to find all the small files. Usually there's a bunch of small temp files lying around if its a email or spooler.

ChooseLife

ChooseLife wrote: »

Question: Suppose you are on a production server and are trying to create a new log file. However, you encounter the following behaviour:

$ echo Test > /var/log/my_new_custom.log
bash: echo: write error: No space left on device
$ su
# echo Test > /var/log/my_new_custom.log
bash: echo: write error: No space left on device

...
Why is the system complaining about lack of space? What command can you execute to find the root cause of the issue?

This question got a fair share of responses (some via PM), so I'll post the answer.

The system is giving a "No space left on device" error because all of the inodes for the partition have been used up.
A quick way to confirm that is to do "df -i", which shows inode usage per partition, and then dig deeper by counting files in directories to discover where the inodes are being used up.

ChooseLife

onesaint wrote: »

I've got (maybe an easy) one for you all:

Let's say I accidentally did an rm -rf on ../httpd/access_log but I still need the file.

Is it recoverable? If so, how?

Very good question. What filesystem are we talking about here?

onesaint

Great question and answer, ChooseLife.

To answer your question, my solution is from systems with ext3/4, however the solution is available across many distros making it FS independent.

W Stewart

Not sure if this question has been asked but I'll go ahead and ask it.

I've got a 3TB harware RAID 1 volume. I can view the drive with

parted -l

but not with

fdisk -l

. Why is that?

Clevernamehere

I'm thinking something to do with fdisk not being able to create partitions over 2TB?