diff options
author | Bryan Newbold <bnewbold@archive.org> | 2021-09-30 16:49:47 -0700 |
---|---|---|
committer | Bryan Newbold <bnewbold@archive.org> | 2021-09-30 16:49:47 -0700 |
commit | f6125848f627ae9bfd3a36d807d2349e1c66bfe3 (patch) | |
tree | 8419947e17dd27621a5b17098eacaedf1a0227a5 /sql/reingest_quarterly.sh | |
parent | 24edc1c754d35df1020907a9c8941f16eaf71440 (diff) | |
download | sandcrawler-f6125848f627ae9bfd3a36d807d2349e1c66bfe3.tar.gz sandcrawler-f6125848f627ae9bfd3a36d807d2349e1c66bfe3.zip |
refactor reingest scripts
Diffstat (limited to 'sql/reingest_quarterly.sh')
-rwxr-xr-x | sql/reingest_quarterly.sh | 17 |
1 files changed, 8 insertions, 9 deletions
diff --git a/sql/reingest_quarterly.sh b/sql/reingest_quarterly.sh index 2e19f5e..20fd82b 100755 --- a/sql/reingest_quarterly.sh +++ b/sql/reingest_quarterly.sh @@ -7,14 +7,13 @@ set -o pipefail # fail if part of a '|' command fails sudo -u postgres psql sandcrawler < dump_reingest_quarterly.sql cd ../python -sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_spn2-error_current.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_spn2-error_current.json -sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_current.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_current.json -#sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_bulk_current.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_bulk_current.json -sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_wayback-error_current.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_wayback-error_current.json -sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_gateway-timeout.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_gateway-timeout.json -sudo -u sandcrawler pipenv run ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_petabox-error_current.rows.json | shuf > /srv/sandcrawler/tasks/reingest_quarterly_petabox-error_current.json +sudo -u sandcrawler pipenv run \ + ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/reingest_quarterly_current.rows.json \ + > /srv/sandcrawler/tasks/reingest_quarterly_current.json -cat /srv/sandcrawler/tasks/reingest_quarterly_spn2-error_current.json /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_current.json /srv/sandcrawler/tasks/reingest_quarterly_wayback-error_current.json /srv/sandcrawler/tasks/reingest_quarterly_petabox-error_current.json /srv/sandcrawler/tasks/reingest_quarterly_gateway-timeout.json | shuf | head -n250000 | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests-daily -p -1 - -#cat /srv/sandcrawler/tasks/reingest_quarterly_cdx-error_bulk.json | shuf | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests-bulk -p -1 +cat /srv/sandcrawler/tasks/reingest_quarterly_current.json \ + | shuf \ + | head -n120000 \ + | jq . -c \ + | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests-daily -p -1 |