aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--sql/dump_reingest_quarterly.sql4
-rw-r--r--sql/dump_reingest_weekly.sql4
-rwxr-xr-xsql/reingest_quarterly.sh6
-rwxr-xr-xsql/reingest_weekly.sh6
4 files changed, 14 insertions, 6 deletions
diff --git a/sql/dump_reingest_quarterly.sql b/sql/dump_reingest_quarterly.sql
index 303824b..066df9f 100644
--- a/sql/dump_reingest_quarterly.sql
+++ b/sql/dump_reingest_quarterly.sql
@@ -10,6 +10,10 @@ COPY (
AND ingest_file_result.status like 'spn2-%'
AND ingest_file_result.status != 'spn2-error:invalid-url-syntax'
AND ingest_file_result.status != 'spn2-error:filesize-limit'
+ AND ingest_file_result.status != 'spn2-error:not-found'
+ AND ingest_file_result.status != 'spn2-error:blocked-url'
+ AND ingest_file_result.status != 'spn2-error:too-many-redirects'
+ AND ingest_file_result.status != 'spn2-error:network-authentication-required'
AND ingest_file_result.status != 'spn2-wayback-error'
) TO '/grande/snapshots/reingest_quarterly_spn2-error_current.rows.json';
diff --git a/sql/dump_reingest_weekly.sql b/sql/dump_reingest_weekly.sql
index 28547a4..bc10b22 100644
--- a/sql/dump_reingest_weekly.sql
+++ b/sql/dump_reingest_weekly.sql
@@ -10,6 +10,10 @@ COPY (
AND ingest_file_result.status like 'spn2-%'
AND ingest_file_result.status != 'spn2-error:invalid-url-syntax'
AND ingest_file_result.status != 'spn2-error:filesize-limit'
+ AND ingest_file_result.status != 'spn2-error:not-found'
+ AND ingest_file_result.status != 'spn2-error:blocked-url'
+ AND ingest_file_result.status != 'spn2-error:too-many-redirects'
+ AND ingest_file_result.status != 'spn2-error:network-authentication-required'
AND ingest_file_result.status != 'spn2-wayback-error'
) TO '/grande/snapshots/reingest_weekly_spn2-error_current.rows.json';
diff --git a/sql/reingest_quarterly.sh b/sql/reingest_quarterly.sh
index 44a22b3..8b0889d 100755
--- a/sql/reingest_quarterly.sh
+++ b/sql/reingest_quarterly.sh
@@ -9,12 +9,12 @@ sudo -u postgres psql sandcrawler < dump_reingest_quarterly.sql
cd ../python
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_spn2-error_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_spn2-error_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_cdx-error_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_cdx-error_current.json
-pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_cdx-error_bulk_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_cdx-error_bulk_current.json
+#pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_cdx-error_bulk_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_cdx-error_bulk_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_wayback-error_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_wayback-error_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_gateway-timeout.rows.json | shuf > /grande/snapshots/reingest_quarterly_gateway-timeout.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_quarterly_petabox-error_current.rows.json | shuf > /grande/snapshots/reingest_quarterly_petabox-error_current.json
-cat /grande/snapshots/reingest_quarterly_spn2-error_current.json /grande/snapshots/reingest_quarterly_cdx-error_current.json /grande/snapshots/reingest_quarterly_wayback-error_current.json /grande/snapshots/reingest_quarterly_petabox-error_current.json /grande/snapshots/reingest_quarterly_gateway-timeout.json | shuf | head -n100000 | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests -p -1
+cat /grande/snapshots/reingest_quarterly_spn2-error_current.json /grande/snapshots/reingest_quarterly_cdx-error_current.json /grande/snapshots/reingest_quarterly_wayback-error_current.json /grande/snapshots/reingest_quarterly_petabox-error_current.json /grande/snapshots/reingest_quarterly_gateway-timeout.json | shuf | head -n250000 | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests -p -1
-cat /grande/snapshots/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 /grande/snapshots/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
diff --git a/sql/reingest_weekly.sh b/sql/reingest_weekly.sh
index dfd4869..8710f76 100755
--- a/sql/reingest_weekly.sh
+++ b/sql/reingest_weekly.sh
@@ -9,12 +9,12 @@ sudo -u postgres psql sandcrawler < dump_reingest_weekly.sql
cd ../python
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_spn2-error_current.rows.json | shuf > /grande/snapshots/reingest_weekly_spn2-error_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_cdx-error_current.rows.json | shuf > /grande/snapshots/reingest_weekly_cdx-error_current.json
-pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_cdx-error_bulk_current.rows.json | shuf > /grande/snapshots/reingest_weekly_cdx-error_bulk_current.json
+#pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_cdx-error_bulk_current.rows.json | shuf > /grande/snapshots/reingest_weekly_cdx-error_bulk_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_wayback-error_current.rows.json | shuf > /grande/snapshots/reingest_weekly_wayback-error_current.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_gateway-timeout.rows.json | shuf > /grande/snapshots/reingest_weekly_gateway-timeout.json
pipenv run ./scripts/ingestrequest_row2json.py /grande/snapshots/reingest_weekly_petabox-error_current.rows.json | shuf > /grande/snapshots/reingest_weekly_petabox-error_current.json
-cat /grande/snapshots/reingest_weekly_spn2-error_current.json /grande/snapshots/reingest_weekly_cdx-error_current.json /grande/snapshots/reingest_weekly_wayback-error_current.json /grande/snapshots/reingest_weekly_petabox-error_current.json /grande/snapshots/reingest_weekly_gateway-timeout.json | shuf | head -n40000 | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests -p -1
+cat /grande/snapshots/reingest_weekly_spn2-error_current.json /grande/snapshots/reingest_weekly_cdx-error_current.json /grande/snapshots/reingest_weekly_wayback-error_current.json /grande/snapshots/reingest_weekly_petabox-error_current.json /grande/snapshots/reingest_weekly_gateway-timeout.json | shuf | head -n60000 | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests -p -1
-cat /grande/snapshots/reingest_weekly_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 /grande/snapshots/reingest_weekly_cdx-error_bulk.json | shuf | jq . -c | kafkacat -P -b wbgrp-svc263.us.archive.org -t sandcrawler-prod.ingest-file-requests-bulk -p -1