Searching Errors in a log file

[Shivkumar]

Hi,

I am looking out for a shell script to accomplish the below :-

We want to capture each occurance of the word Error ( case insensitive ) in a log file and 5 lines below each of the lines and store the output in a file.

Can anyone post the script ?

Thanks,
Shiv

[Peter Nikitka · Accepted solution]

Hi,

Some open issues:
- you say 'word': is it really a separate word (white space around)? 'Error:' for example would NOT meet this criterium.
- you say case insensitive: may the be ANY combinations of Error, ErrOR or is there a restriction to Error, error and ERROR?

I offer an awk solution (untested).

awk '/Error/ || /error/ || /ERROR/ {count=6}
count {print;count--}' infile >outfile

mfG Peter

[john korterman]

Hi,

try the attached script with these parameters: $1="Error" , $2="5", $3="path_to_file"

regards,
John K.

[Hein van den Heuvel]

Perl regular expression have something a 'break' match between words: \b

It helps to make sure that a script like requested does trigger on "Error:" but not on "terrorist".

Example usage:

perl -ne "$count=5 if /\berror\b/i; print if $count-- > 0" your-log-file

hth,
Hein.

[Bill Hassell]

And just a note about other 'error' conditions. You might want to search for other keywords that also signify problems:

grep -i -e fail -e warn -e alert -e critical -e abort /some-dir/some-log

The above grep just shows the line where one or more of the search strings match. Use it to see if there are any other messages that might be important but do not have Error in the entry.

[James R. Ferguson]

Hi Shiv:

Bill makes a good point, and we can easily use Hein's Perl snippet thusly:

# perl -ne '$count=5 if /\b(error|warn|critical|alert|abort|fail)\b/i;print if $count-- > 0' logfile

Thus if the file (or files) you specify as arguments contain any of the words "error", "warn", "critical", "alert", "abort" or "fail", the line on which any one occurs will be printed and four more lines afterwards.

As written, you can process multiple files at one simply by passing their names as arguments. If you would like to do this, *and* keep track of the filename for which matches occur, modify the script thusly:

# perl -ne '$count=5 if /\b(error|warn|critical|alert|abort|fail)\b/i;print "$ARGV:$_" if $count-- > 0' log1 log2 log3

...Now output can look like:

log1:critical alert: line-3
log1:line-4
log1:line-5
log1:line-6
log1:line-7
log2:ERROR at line-6
log2:line-7
log2:line-8
log2:line-9
log2:line-10

Regards!

..JRF...

[Rory R Hammond]

This AWK example works hpux 11.11
I put a space before each word.
so it would get things like " Error:"

awk -F: '
(tolower($0) !~ / error| warn| alert| critical| abort| invalid/) {next}
{
print "\n" $0
for (i=1;i < 6;i++)
{getline
print $0
}
} ' /var/adm/syslog/syslog.log

[James R. Ferguson]

Hi Rory:

Yes, placing a space (blank) character before the words in the 'awk' solution helps, but it dones *not* solve the number of matches that Perl does.

For starters, what if the space were a tab character? The use of '\b' in Perl regular expressions covers word boundries delimited by spaces, tabs and most punctuation characters like ":" and "-" that might appear like "alert-bad thing happened".

Regards!

...JRF...

[Coolmar]

Hi,

This may not be what you require, but I have used it alot. It will watch your log files and email you whenever it finds any security violations or errors. YOu can tailor it to watch for certain words - like "error"

http://sourceforge.net/projects/logcheck/

[Rory R Hammond]

Jim,

Like all programming problems there are many tools with many solutions. My intent was to show how you can use grep like expressions inside of awk The tolower/toupper function inside the if statement is novel approach, along with the regular expression syntax. Of course, I made the fatal error of assuming that the only type of white space is a "space". Which in the syslogs I looked at was true.

a partial solution:

awk -F: '
(tolower($0) ~ /[ \t]error|[ \t]warn|[ \t]alert|[ \t]critical|[ \t]abort|[ \t]invalid/ ) {
print "\n" $0
for (i=1;i < 6;i++)
{
getline
print $0
}
} ' /var/adm/syslog/syslog.log

To bad I could not get the following to work:
(tolower($0) ~ /[:space:]error|[:space:]warn|[:space:]alert|[:space:]critical|[:space:]abort|[:space:]invalid/)

Rory

[James R. Ferguson]

Hi Rory:

I did not mean any offense; rather only that Perl has some of the best regular expression support available. Your use of '[:space:]' enables whitespace (a space or a tab) detection and is good defensive technique. This works:

# awk '{if (tolower($0) ~/[[:space:]]error|[[:space:]]warn/) {print $0}}' filename

Note the double square brackets. This is necessary to create a character class consisting of '[:space:]'.

Regards!

...JRF...

[Rory R Hammond]

Jim,

No offense was taken and I hope my reply did not suggest that I was offended. The discussion about problems and solutions are always good.


Rory

[Shivkumar]

Hi John,

Is this the right syntax to use your script ?

$scriptname.sh "Error" "5" "logfilename"

[Shivkumar]

I was trying to execute the "john korterman's" script. I got the error message.

./tlog: syntax error at line 25: `TOTAL_LINES=$' unexpected

There appears to be some syntax error in the line:-
TOTAL_LINES=$(awk '{nlines = nlines + 1}' END {print nlines} <$3)

=====================================================================
Below is the complete script.

script name:tlog

#!/usr/bin/sh
# Print number of lines, $2,
# after matched , case-insensitive string, $1,
# from file, $3
#

typeset -i FIRST_LINE=0 LAST_LINES=0 MATCH_LINE_NO=0 LINE_AFTER=0 TOTAL_LINES=0


# Check params..
if [ "$#" != 3 ]
then
echo wrong number of parameters
echo par1=string, par2=number of lines, par3=file
exit 1
fi
#
if [ ! -r $3 ]
then
echo Cannot read $3
exit 1
fi

# Number of lines in $3
TOTAL_LINES=$(awk '{nlines = nlines + 1}' END {print nlines} <$3)



grep -in "$1" "$3" | while read LINE
do
# Line number for match
MATCH_LINE_NO=$(echo $LINE | awk -F: '{print $1}')

# Check line number boundaries
FIRST_LINE=$(($MATCH_LINE_NO))
#
LAST_LINE=$(($MATCH_LINE_NO + $2))
if [ "$LAST_LINE" -gt "$TOTAL_LINES" ]
then
echo match in line $MATCH_LINE_NO, but cannot print lines after $TOTAL_LINES
fi

# Print after...
LINE_AFTER=$(($MATCH_LINE_NO + $2))
cat -n $3 | awk -v lineb=$FIRST_LINE -v linea=$LINE_AFTER '$1==lineb,$1==linea {print $0}'
echo
done
====================================================================
Appreciate suggestion.

Thanks,
Shiv

[Sandman!]

You are missing the terminating single-quote for the awk construct. The single quote should be placed after the action specified by the END statement...imho change the line below:

from...
TOTAL_LINES=$(awk '{nlines = nlines + 1}' END {print nlines} <$3)
to...
TOTAL_LINES=$(awk '{nlines = nlines + 1} END {print nlines}' <$3)

~hope it helps

[Shivkumar]

Hi Sandman,

now i am getting different error as:-
`MATCH_LINE_NO=$' unexpected

for the below line in the script:-

# Line number for match
MATCH_LINE_NO=$(echo $LINE | awk -F: '{print $1}')

[James R. Ferguson]

Hi Shiv:

I'm not sure what you have. I copy-and-pasted John's script from your post of today (Dec 20). I corrected line-25 as Sandman pointed out and can run the script without errors and receive the output I expect.

I suggest you re-copy-and-paste and try the same.

Regards!

...JRF...

[Sandman!]

Hi Shiv,

I copied John's script and it runs perfectly on my machine. As JRF says there is a problem with the copy 'n paste on your end.

~cheers

[Shivkumar]

Now getting new errors.

Here is the messages:-

1st test:

$./t1 "Errors" 5 logfile
awk: record `java.class.path = /o...' too long
record number 1653
match in line 1732, but cannot print lines after 0
awk: syntax error near line 1
awk: bailing out near line 1


2nd test:
$./t1 "listening on port 1000" 5 logfile

match in line 2013, but cannot print lines after 0
awk: syntax error near line 1
awk: bailing out near line 1

Looks like some awk syntax errors ??

[James R. Ferguson]

Hi Shiv:

From your output, I would guess that the log file has a record (a string of characters delimited with a newline) that exceeds an 'awk' internal limit. You might find, too, that you can't 'vi' the file for the same reason. Perl has no practical limits.

Regards!

...JRF...