Mercurial > hg > cc > cirrus_home

annotate bin/doExtract.sh @ 166:afd7879181c9

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
old style
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Mon, 18 Jul 2022 19:15:20 +0100
parents ef3533d3ac4b
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
27
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
1 #!/bin/bash
29
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
2 # Usage doExtract.sh ccid jobid segid
27
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
3 ccid=$1
29
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
4 export jobid=$2
27
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
5 segid=$3
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
6 echo $(date) $(hostname) $jobid $segid
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
7 mkdir -p $segid/logs
7a2bc060230d first cut at doing extraction here
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
8 cd $segid
29
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
9 ls /beegfs/common_crawl/CC-MAIN-${ccid}/${segid}/CC-MAIN-*.warc.gz | \
33
4c23b1766692 towards sub-division of resulting tar files
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 30
diff changeset
10 parallel --joblog job_${jobid}.log -j 18 -N 1 'id=$(echo {} | cut -f 6 -d / | cut -f 3- -d - | cut -f 1 -d .) ; \
35
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
11 fid=$(printf "%03.0f" $(({#} - 1))) ; \
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
12 lf=logs/${jobid}_${fid}_log ; \
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
13 echo starting $id $(date) >> $lf ; \
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
14 unpigz -dp 1 -c {} | $HOME/lib/valhalla/bin/warc.sh ${id} application/pdf 2>> $lf ; \
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
15 echo finished ${id} $(date) >> $lf'
29
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
16 res=$?
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
17 echo $(date) $(hostname) $jobid $segid $res
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
18 if [ $res = 0 ]
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
19 then
35
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
20 pfx=$(ls /beegfs/common_crawl/CC-MAIN-${ccid}/${segid}/CC-MAIN-*-00000.warc.gz |\
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
21 cut -f 6 -d / | cut -f 3,4 -d -)
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
22
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
23 cat ../by11s.txt | while read i j
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
24 do ((n=i/11))
36
e912ed51146a fixed scope pblm in tar step
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 35
diff changeset
25 tar -cf /beegfs/common_crawl/CC-MAIN-${ccid}/${segid}/extract_${n}.tar \
e912ed51146a fixed scope pblm in tar step
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 35
diff changeset
26 $(seq $i $j | xargs -I ^ bash -c '{ k=^; printf "${0}-%05.0f_* logs/*_%03.0f_log\n" $k $k ; }' $pfx)
35
ec99b2d1d2fc sync up filenames and log names,
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 33
diff changeset
27 done &&
37
ef3533d3ac4b clean up after ourselves
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 36
diff changeset
28 echo $(date) $(hostname) $jobid /beegfs/common_crawl/CC-MAIN-${ccid}/${segid}/extracts_\{0..50\}.tar &&
ef3533d3ac4b clean up after ourselves
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 36
diff changeset
29 cd .. &&
ef3533d3ac4b clean up after ourselves
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 36
diff changeset
30 rm -rf $segid
29
a95a1b31f5e9 try to fix multi-line lossage
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 27
diff changeset
31 fi