Re: how to parse oncheck output?

In article <6f6huf$kkt@zax.whoville.leftbank.com>, cos@zax.whoville.leftbank.com (Ofer Inbar) wrote: > > I've written a short script to run all of the consistency checks > recommended in chapter 31 of the Administrator's Guide: > > ---------------------------------------------------------------------- > #!/usr/bin/perl > > open DBLIST, "echo 'select name,owner,created,flags from sysdatabases' > | dbaccess sysmaster - 2>&1 |" > or die "can't open dbacess pipe: $!\\n"; > while (<DBLIST>) { push @dblist, $1 if > /^(\\w+)\\s+\\S+\\s+\\d\\S+\\d\\s+\\d+\\s*$/ } > close DBLIST; > > print `oncheck -cr`; > print `oncheck -ce`; > print `oncheck -cc`; > > foreach $db (@dblist) > { print `oncheck -cI $db`; > print `oncheck -cD $db`; > } > ---------------------------------------------------------------------- > > What I'd like to do is run it from crontab and have it mail its output > to a few people. But, oncheck is really verbose! I'd like it to send > mail only if there's an actual problem that someone needs to worry > about. It would be trivial for me to change those backtick calls into > opens of pipes, and have perl parse the output to figure out if it's > worth mailing out. Unfortunately, I have no clue what to look for. > There doesn't seem to be any regular format for what oncheck spews > out, nor is the format of its messages and error indications > documented anywhere that I can find. > > One possible rule might be: > Look for the word "error". Starting with the line that has the word > "error", continue until you reach a blank line. That block of text is > the entire error message for an actual problem. > > If that were true, I could write my script accordingly. It seems like > a plausible rule to me, based on what I've seen of oncheck output so > far, but without documentation I just don't *know*. I don't want to > make a guess that will end up suppressing actual error messages that > we need to see, leading to a false sense of confidence. But I also > don't want to mail out the entire output every day, because people > will stop reading it. > > I posted this question once before, about a year ago. I got one > followup, but it was a big .shar of a bunch of shell & sed scripts > that did a whole bunch of things not directly related to my question, > and it was written for and tested Informix 5 whereas I'm running 7.22 > > I don't actually want any scripts, because I want to tailor my script > to our local needs anyway, and I definitely want to keep it all in > perl. All I want is some documentation, or pointer to documentation, > or description, of all the kinds of output oncheck might produce and > how an automated script is supposed to be able to figure out what is > important to report about. Since our database is mostly fine, and has > always been so, I've never actually SEEN any real errors reported by > oncheck, so I have no idea what they would look like when they do > appear. I want to make sure my script will be able to handle anything > oncheck throws at it. For example, being told "a serious error will > aways start with the word 'ERROR:'", in and of itself, isn't helpful > because perhaps some errors give multiline output, and I'd need to > know how to identify those, and how to identify where the error > message ends. > > I'm not afraid of reading manuals, but I just can't find this > information and I don't know where to look for it. > > -- Cos (Ofer Inbar) -- cos@leftbank.com cos@cs.brandeis.edu > -- The Left Bank Operation -- lbo@leftbank.com http://www.leftbank.com/ > It's been said that if a sysadmin does his job perfectly, he's the > fellow that people wonder what he does and why the company needs him, > until he goes on vacation. -- comp.unix.admin FAQ > I use the output of oncheck -pT a lot. I have found a shell script ages ago in the IIUG archives (look at www.iiug.org, but you know this anyway) which brought tablename, allocated / used pages, first/next extentsize into one line of text. I then modified this a lot (V5->V7...) and now it gives me a file which I can dbload into a table. From there I collet grow rate in extents/pages/rows per server and month/week (we have 99 machines), actual fill grade of indices & more information of that sort The sum function gives me an forecast how long it will take onarchive to perform a LEVEL 0 ARCHIVE. I am also looking for more information explaining all the output. (oncheck -p{kKlL}) Out oncheck -cc always gives warnings so plus to the '-q' switch I parse it's output for the word 'ERROR/error' as '-q' doesn't prevent all the warnings & I can live with that (Never saw an error) I am especially interested to learn if somebody has written a script dealing with getting out which tables are to be reorganized TOGETHER because we see extents of 10-15 tables lying around so, that to be able to recreate the tables creating no 'holes' on disk & with bigger first extent size we first have to copy/remove a buch of tables extents to create sufficient contiguous space to allocate the then bigger pieces. This is a disk defrag problem - any solutions ??? We use SNI RM400/RM600 & SINIX V5.43 & INFORMIX V7.12 UC1 Regards Dick Richard Kofler EDVg-debis Systemhaus Vienna, Austria e-mail: richard_kofler@edvg.co.at -----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/ Now offering spam-free web-based newsreading