Mercurial > hg > cc > azure

comparison workers/bin/timedWhich.sh @ 17:2a2c1fb03c54

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first cut at http/https real trial, with month and year last-modified info too
author Henry S. Thompson <ht@markup.co.uk>
date Fri, 19 Oct 2018 11:36:31 +0000
parents
children 9631fca89cc6
comparison
equal deleted inserted replaced
16:c3e9ad8a42cf 17:2a2c1fb03c54
1 #!/bin/bash
2 # Test script to split CC WAT files across threads
3 # to tabulate http vs. https by last-modified date:
4 # Usage: [echo file file_id] | timedWhich.sh id home [-t] numWorkerProcesses
5 # If -t, no random wait, just id seconds
6 # remove >>errs once tested
7 #set -e -o pipefail
8 echo $$ > test1.pid
9 proc=$1
10 res=res$proc
11 home=$2
12 shift 2
13 function lrand {
14 # cheap bad little random number generator
15 echo $(( 1 + ($(openssl rand 1 | od -d | head -1 | tr -s ' ' | cut -f 2 -d ' ') % $1)))
16 }
17 if [ "$1" = "-t" ]
18 then
19 shift
20 pause=$proc
21 else
22 pause=$(lrand 60)
23 fi
24 wp=$1
25 touch .running
26 function tryRead {
27 m=0
28 set -o pipefail
29 until if [ $((m+=1)) -gt 5 ]; then echo "tried to read $1 5 times w/o success, giving up" 1>&2; return 1; fi && \
30 curl -s --insecure -o - "$1"| zcat | fgrep msgtype=response | \
31 { egrep -i '"WARC-Target-URI":"https?:' || : ;} > $2
32 do
33 # try to avoid lockstep retries
34 echo ${PIPESTATUS[@]} 1>&2
35 sleep $(lrand 10)
36 echo $(date) $2 retry number $m 1>&2
37 done
38 set +o pipefail
39 }
40 trap "{
41 #set -e -o pipefail
42 ln -s ../nohup.cc .
43 tar -czhf - CC* $res | \
44 ssh -o StrictHostKeyChecking=no -q $home \"{ cd data
45 mkdir -p pdf/wat
46 cd pdf/wat
47 tar -xzf - ; } 2>>errs\"
48 rm -rf $res CC* ifile.txt *.pid
49 ( sleep 5 ; rm nohup.cc ) &
50 }" EXIT
51 mkdir -p $res
52 log=$res/log
53 # Don't all start at once
54 sleep $pause
55 echo \# $(date) > $log
56 pRes=0
57 while read s id
58 do
59 url="https://commoncrawl.s3.amazonaws.com/$s"
60 ccm=${s##*/wat/}
61 cci=${ccm%%-ip*}
62 export ID=$id
63 echo \# $(date) "running |$proc|$home|$pause|$wp|$id|" >> $log
64 # Experimental retry loop
65 tryRead "$url" crawl$id
66 if [ -s crawl$id ]
67 then
68 echo \# $(date) $id $(wc -l crawl$id) >> $log
69 parallel --round-robin --pipe --block-size 2M -j $wp "_timedWhich.sh {#} >> $res/$cci.{#} 2>>$res/errs{#}" < crawl$id || echo "ppfailed $? ${PIPESTATUS[@]}" 1>&2
70 else
71 echo "crawl$id empty" 1>&2
72 fi
73 rm crawl$id
74 cat $res/$cci.* > $cci
75 done < ifile.txt 2>> $res/errs || pRes=$?
76 echo \# $(date) main loop exit code=$pRes >> $log
77 rm .running
78
79