PTS system uses "meta" table which contains a single row.
This table gets updated every time PTS record is updated.
If you have many processes that want to manage PTS (e.g. 100s or 1000s of changefeeds)
all of these changefeeds will contend against single row update.
This does not scale, and causes long running retries/backoffs.
PTS system uses "meta" table which contains a single row. This table gets updated every time PTS record is updated. If you have many processes that want to manage PTS (e.g. 100s or 1000s of changefeeds) all of these changefeeds will contend against single row update. This does not scale, and causes long running retries/backoffs.
Jira issue: CRDB-31059
gz#18553
Epic CRDB-34798
gz#20330