aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBryan Newbold <bnewbold@archive.org>2022-01-05 11:54:49 -0800
committerBryan Newbold <bnewbold@archive.org>2022-01-05 11:54:49 -0800
commit65a0d38bcedca0610ca6fa8e053199f324062ace (patch)
treecb210178e213f0384698df38f02c862795cc2263
parente4c153d4f6d8c815c7e792fec2f40cdac39bab35 (diff)
downloadsandcrawler-65a0d38bcedca0610ca6fa8e053199f324062ace.tar.gz
sandcrawler-65a0d38bcedca0610ca6fa8e053199f324062ace.zip
document progress on re-GROBID-ing
-rw-r--r--notes/tasks/2021-12-06_regrobid.md89
1 files changed, 89 insertions, 0 deletions
diff --git a/notes/tasks/2021-12-06_regrobid.md b/notes/tasks/2021-12-06_regrobid.md
index 65e9fe3..d879277 100644
--- a/notes/tasks/2021-12-06_regrobid.md
+++ b/notes/tasks/2021-12-06_regrobid.md
@@ -191,6 +191,82 @@ And some earlier files of interest on `aitio`:
| pv -l \
| kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.unextracted -p -1
+
+## Ancient Fatcat Files
+
+Files from an era where we didn't record GROBID version or status, even for
+success.
+
+ COPY (
+ SELECT DISTINCT ON (cdx.sha1hex) row_to_json(cdx)
+ FROM grobid
+ LEFT JOIN cdx ON grobid.sha1hex = cdx.sha1hex
+ LEFT JOIN fatcat_file ON grobid.sha1hex = fatcat_file.sha1hex
+ WHERE
+ grobid.status_code = 200
+ AND grobid.status IS NULL
+ AND cdx.sha1hex IS NOT NULL
+ AND fatcat_file.sha1hex IS NOT NULL
+ -- sort of arbitary "not recently" date filter
+ AND (grobid.updated IS NULL OR grobid.updated < '2021-11-15')
+ -- LIMIT 5;
+ )
+ TO '/srv/sandcrawler/tasks/ungrobided_fatcat.2021-12-11.grobid_status_null.json'
+ WITH NULL '';
+
+ cat /srv/sandcrawler/tasks/ungrobided_fatcat.2021-12-11.grobid_status_null.json \
+ | rg -v "\\\\" \
+ | jq . -c \
+ | pv -l \
+ | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ungrobided-pg -p -1
+ # 107k 0:00:03 [29.9k/s]
+
+
+## Start Re-Processing Old GROBID Versions
+
+ COPY (
+ SELECT DISTINCT ON (cdx.sha1hex) row_to_json(cdx)
+ FROM grobid
+ LEFT JOIN cdx ON grobid.sha1hex = cdx.sha1hex
+ LEFT JOIN fatcat_file ON grobid.sha1hex = fatcat_file.sha1hex
+ WHERE
+ grobid.status = 'success'
+ AND grobid.grobid_version NOT LIKE '0.7.%'
+ AND cdx.sha1hex IS NOT NULL
+ AND fatcat_file.sha1hex IS NOT NULL
+ -- sort of arbitary "not recently" date filter
+ AND (grobid.updated IS NULL OR grobid.updated < '2021-11-15')
+ -- LIMIT 5;
+ )
+ TO '/srv/sandcrawler/tasks/ungrobided_fatcat.2021-12-11.grobid_old.json'
+ WITH NULL '';
+
+This one is huge, and want to process in batches/chunks of ~8 million at a time.
+
+ cd /srv/sandcrawler/tasks/
+ cat /srv/sandcrawler/tasks/ungrobided_fatcat.2021-12-11.grobid_old.json \
+ | split --lines 5000000 - ungrobided_fatcat.2021-12-11.grobid_old.split_ -d --additional-suffix .json
+
+Submit individual batches like:
+
+ cat /srv/sandcrawler/tasks/ungrobided_fatcat.2021-12-11.grobid_old.split_01.json \
+ | rg -v "\\\\" \
+ | jq . -c \
+ | pv -l \
+ | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ungrobided-pg -p -1
+
+Overall progress:
+
+ x ungrobided_fatcat.2021-12-11.grobid_old.split_00.json
+ . ungrobided_fatcat.2021-12-11.grobid_old.split_01.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_02.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_03.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_04.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_05.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_06.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_07.json
+ => ungrobided_fatcat.2021-12-11.grobid_old.split_08.json (small)
+
## General Counts
How many fatcat files of what mimetype (reported in sandcrawler-db)?
@@ -287,3 +363,16 @@ What are the GROBID status codes for fatcat files? Narrowed down:
error | 200 | 3
(7 rows)
+Ran the same query again on 2021-12-15:
+
+ status | status_code | count
+ ----------------+-------------+----------
+ success | 200 | 45092915
+ error | 500 | 302373
+ | | 250335
+ | 200 | 53352
+ bad-grobid-xml | 200 | 39
+ error-timeout | -4 | 37
+ error | 200 | 34
+ error | 503 | 2
+ (8 rows)