UNIX for Dummies Questions & Answers

Hi,
I have a question, during my readings it appears that these two variables in the snippet below need to be on the same line to return a “true” answer and listed in the output otherwise it won’t be returned. How can I write this snippet to make it return a “true” if those two variables are on two separate lines in the trace file?? How would this piece of the code be written?

Code:

egrep  -ce  "($2015-01-22.*$NS Primary Error) | ($NS Primary Error.*$2015-01-22)" /oracle/diag/rdbms/musidp/musidp/trace/musidp_d000_21750.trc

Or am I misunderstanding something here?
Thanks in advance.

is your $NS variable set?
Or better yet... what's this $2015-01-22? Looks like a variable called 2015-01-22
Shouldn't this simply be:

Code:

egrep  -ce  '(2015-01-22|NS Primary Error)' /oracle/diag/rdbms/musidp/musidp/trace/musidp_d000_21750.trc

First off, your regex is a little strange. "$2015-01-22" is nonsensical because $ means "match the end of the line", and there's nothing after the end of the line. Did you mean ^, "match the beginning of the line"? It will also take your spaces literally, so "(...) | (...)" means "match this block and a space, or match a space then this block".

But anyways, grep can't can't remember what happened in past lines. It's not a programming language.

awk can remember what's happened in past lines, because it is.

Code:

awk '/2015-01-22.*NS Primary Error/ { A=1 } /NS Primary Error.*2015-01-22/ { B=1 } END { exit(!(A && B)); }' inputfile

If it finds lines matching "2015-01-22.*NS Primary Error" it sets the variable A true. If it finds lines matching "NS Primary Error.*2015-01-22" it sets the variable B true. If, once all lines are read, both A and B are true, it returns 0(i.e. success), otherwise returns 1 (failure) to the shell.

You might list the lines of each pattern in different work files a and b using 2 egrep patterns/passes, and then sort each (comm needs a binary not ascii-numeric sort) and use 'comm' -3 to remove the same-line entries. What comes out is either a file lines or b file lines prefixed with a tab.

Hi Vgers
Have to admit I am really new at this and am trying to learn by reading and doing.

The code that I listed was wrong I was just trying something. I tried your code and it returned a number (29). I went into the file to verify if it returned the right number but it didn’t. I counted 5 2015-01-22 instances of this search string and 24 instances of the NS primary error string. See output below.
What I need to have is to return true only that number that has BOTH the date and the NS primary error string which the answer should be 5.
I am thinking that there should be two egreps on for 2015-01-22 and then egrep for NS primary error.
Here is the data that egrep is looking for together (highlighted) snippet from trace file

Code:

*** 2015-01-22 17:13:33.267
unexpected error 12560 for connection:
NS Primary Error: TNS-12535: TNSperation timed out
NS Secondary Error: TNS-12560: TNSrotocol adapter error
NT Generic Error: TNS-00505: Operation timed out

Both of these highlighted must return true to be counted.
My updated script from your suggestion does this:

Code:

egrep -e '.*2015-01-22' /oracle/diag/rdbms/musidp/musidp/trace/musidp_d000_21750.trc | egrep -ce 'NS Primary Error' '{}' '+'

The first half of the script returns the following so when that info gets to the second egrep it will always return zero.
RETURNS THE FOLLOWING OUTPUT:

Code:

2015-01-22 11:21:29
2015-01-22 11:29:31
2015-01-22 11:32:01
2015-01-22 11:34:22
2015-01-22 11:35:16

Right now I am stumped. I hope this additional information helps.
Thanks in advance
al

---------- Post updated at 07:03 PM ---------- Previous update was at 06:43 PM ----------

Hi Corona,

Thanks for taking a look at this for me. I am really new at this and am currently learning on the job.

I ran your statement on the server and it returned the following error:

Code:

awk: syntax error near line 1
awk: illegal statement near line 1

I ran it with just one trace file with known information in it. In the end though I will need to go through the entire directly of trace files to find the appropriate ones.

Regards,
al

On Solaris use /usr/xpg4/bin/awk rather than awk

--
Your egrep approach will not work, because it is line based. The syntax of the second part is incorrect too and it the regular expression on the first part can be reduced to

Code:

egrep 2015-01-22

In the first example you use a -c option which counts the number of appearances. But now it is the actual result you are after?

Please provide a corresponding input sample and output sample and also verbally describe what result you are trying to achieve. It also helps if you state your specific operating system and version. I guessed it this time from the error message you reported....

---
Also, to find al matches, try:

Code:

/usr/xpg4/bin/awk -v ts="2015-01-22" '/^\*\*\*/ {if($0~ts) p=$0; else p=x} p!="" && /Primary Error/{print p ORS $0}' file

Or, to combine this with count and exit status, like Corona688 suggested, try something a bit more elaborate:

Code:

/usr/xpg4/bin/awk -v ts="2015-01-22" '
  /^\*\*\*/ {
    if($0~ts) p=$0
    else p=x
  } 
  p!="" && /Primary Error/ {
    print p ORS $0
    c++
  } 
  END { 
    print c+0 " Matches found"
      if(c==0) exit 1
  }
' infile

Yes, even 'sed' is more capable of dealing with multiple lines, using 'N' to pile up the entre block for testing, perhaps removing all but interblock line feeds.