diff options
Diffstat (limited to 'notes/tasks')
-rw-r--r-- | notes/tasks/2022-11-21_andrzejklimczuk_cleanup.md | 132 |
1 files changed, 132 insertions, 0 deletions
diff --git a/notes/tasks/2022-11-21_andrzejklimczuk_cleanup.md b/notes/tasks/2022-11-21_andrzejklimczuk_cleanup.md new file mode 100644 index 0000000..74d3857 --- /dev/null +++ b/notes/tasks/2022-11-21_andrzejklimczuk_cleanup.md @@ -0,0 +1,132 @@ + +Had a huge number of SPN requests for the andrzejklimczuk.com domain, +presumably from the author. + +Many were duplicates (same file, multiple releases, often things like zenodo +duplication). Many were also GROBID 500s, due to truncated common crawl +captures. + +Needed to cleanup! Basically sorted through a few editgroups manually, then +rejected all the rest and manually re-submitted with the below queries and +commands: + + SELECT COUNT(*) from ingest_request + LEFT JOIN ingest_file_result ON + ingest_file_result.ingest_type = ingest_request.ingest_type + AND ingest_file_result.base_url = ingest_request.base_url + LEFT JOIN grobid ON + grobid.sha1hex = ingest_file_result.terminal_sha1hex + WHERE + ingest_request.link_source = 'spn' + AND ingest_request.ingest_type = 'pdf' + AND ingest_request.base_url like 'https://andrzejklimczuk.com/%'; + => 589 + + SELECT ingest_file_result.status, COUNT(*) from ingest_request + LEFT JOIN ingest_file_result ON + ingest_file_result.ingest_type = ingest_request.ingest_type + AND ingest_file_result.base_url = ingest_request.base_url + LEFT JOIN grobid ON + grobid.sha1hex = ingest_file_result.terminal_sha1hex + WHERE + ingest_request.link_source = 'spn' + AND ingest_request.ingest_type = 'pdf' + AND ingest_request.base_url like 'https://andrzejklimczuk.com/%' + GROUP BY ingest_file_result.status; + + status | count + ----------------+------- + cdx-error | 1 + success | 587 + wrong-mimetype | 1 + (3 rows) + + + SELECT grobid.status_code, COUNT(*) from ingest_request + LEFT JOIN ingest_file_result ON + ingest_file_result.ingest_type = ingest_request.ingest_type + AND ingest_file_result.base_url = ingest_request.base_url + LEFT JOIN grobid ON + grobid.sha1hex = ingest_file_result.terminal_sha1hex + WHERE + ingest_request.link_source = 'spn' + AND ingest_request.ingest_type = 'pdf' + AND ingest_request.base_url like 'https://andrzejklimczuk.com/%' + GROUP BY grobid.status_code; + + status_code | count + -------------+------- + 200 | 385 + 500 | 202 + | 2 + (3 rows) + + + COPY ( + SELECT row_to_json(ingest_request.*) FROM ingest_request + LEFT JOIN ingest_file_result ON + ingest_file_result.ingest_type = ingest_request.ingest_type + AND ingest_file_result.base_url = ingest_request.base_url + LEFT JOIN grobid ON + grobid.sha1hex = ingest_file_result.terminal_sha1hex + WHERE + ingest_request.link_source = 'spn' + AND ingest_request.ingest_type = 'pdf' + AND ingest_request.base_url like 'https://andrzejklimczuk.com/%' + AND ingest_file_result.status = 'success' + AND grobid.status_code = 500 + ) TO '/srv/sandcrawler/tasks/andrzejklimczuk_bad_spn.rows.json'; + => COPY 202 + + COPY ( + SELECT row_to_json(ingest_request.*) FROM ingest_request + LEFT JOIN ingest_file_result ON + ingest_file_result.ingest_type = ingest_request.ingest_type + AND ingest_file_result.base_url = ingest_request.base_url + LEFT JOIN grobid ON + grobid.sha1hex = ingest_file_result.terminal_sha1hex + WHERE + ingest_request.link_source = 'spn' + AND ingest_request.ingest_type = 'pdf' + AND ingest_request.base_url like 'https://andrzejklimczuk.com/%' + AND ingest_file_result.status = 'success' + AND grobid.status_code = 200 + ) TO '/srv/sandcrawler/tasks/andrzejklimczuk_good_spn.rows.json'; + => COPY 385 + +sudo -u sandcrawler pipenv run \ + ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/andrzejklimczuk_good_spn.rows.json \ + > /srv/sandcrawler/tasks/andrzejklimczuk_good_spn.json + +sudo -u sandcrawler pipenv run \ + ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/andrzejklimczuk_bad_spn.rows.json \ + | jq '. + {force_recrawl: true}' -c \ + > /srv/sandcrawler/tasks/andrzejklimczuk_bad_spn.json + +cat /srv/sandcrawler/tasks/andrzejklimczuk_bad_spn.json \ + | shuf \ + | head -n60000 \ + | jq . -c \ + | kafkacat -P -b wbgrp-svc350.us.archive.org -t sandcrawler-prod.ingest-file-requests-priority -p -1 + +cat /srv/sandcrawler/tasks/andrzejklimczuk_good_spn.json \ + | shuf \ + | head -n100 \ + | jq . -c \ + | kafkacat -P -b wbgrp-svc350.us.archive.org -t sandcrawler-prod.ingest-file-requests-priority -p -1 + +cat /srv/sandcrawler/tasks/andrzejklimczuk_good_spn.json \ + | shuf \ + | head -n10000 \ + | jq . -c \ + | kafkacat -P -b wbgrp-svc350.us.archive.org -t sandcrawler-prod.ingest-file-requests-priority -p -1 + +sudo -u sandcrawler pipenv run \ + ./scripts/ingestrequest_row2json.py /srv/sandcrawler/tasks/andrzejklimczuk_bad_spn.rows.json \ + > /srv/sandcrawler/tasks/andrzejklimczuk_bad2_spn.json + +cat /srv/sandcrawler/tasks/andrzejklimczuk_bad2_spn.json \ + | shuf \ + | head -n60000 \ + | jq . -c \ + | kafkacat -P -b wbgrp-svc350.us.archive.org -t sandcrawler-prod.ingest-file-requests-priority -p -1 |