bootstrap.sh improvements #10490
bootstrap.sh improvements #10490
Conversation
Signed-off-by: Nir Ben-Or <[email protected]>
Codecov ReportAll modified and coverable lines are covered by tests ✅ |
steven-sheehy
added
bug
| # Start a new process group and detach from terminal | ||
| set -m |
There was a problem hiding this comment.
Revert this, it causes problems with spotless.
| @@ -394,7 +398,7 @@ process_manifest() { | |||
|
|
|||
| if [[ -f "$file_path" ]]; then | |||
| # Skip validation if file is already imported successfully | |||
| if grep -q "^$file_path IMPORTED" "$TRACKING_FILE" 2>/dev/null; then | |||
| if [[ "$(read_tracking_status "$(basename "$file_path")")" == "IMPORTED" ]]; then | |||
There was a problem hiding this comment.
read_tracking_status internally calls basename so this can be omitted, I believe.
| current_status=$(read_tracking_status "$file") | ||
| if [[ "$current_status" != "IMPORTED" ]]; then | ||
| write_tracking_file "$file" "IMPORTED" | ||
| fi |
There was a problem hiding this comment.
Does it really matter if you check the status before write? It doesn't harm it to always set status IMPORTED like before.
| fi | ||
| done | ||
|
|
||
| # Check the exit status of the count query |
There was a problem hiding this comment.
If the count query fails after 3 times we should fallback to it returning true after logging an error. In theory, the import should've returned failure if any rows were missed and this is just a safety check. This change alone would've unblocked the customer.
| actual_count=$(psql -v ON_ERROR_STOP=1 -q -Atc "SELECT COUNT(*) FROM ${table};") | ||
| psql_status=$? | ||
| else | ||
| actual_count=$(psql -v ON_ERROR_STOP=1 -q -Atc "SELECT COUNT(*) FROM ${table} WHERE consensus_timestamp BETWEEN '$start_ts' AND '$end_ts';") |
There was a problem hiding this comment.
We could also fallback to a quicker query like where consensus_timestamp in ($start_ts, $end_ts) and verify the first and last are present.
Description:
processing_filesassociative array and relying on the pids array for job trackingvalidate_special_filesfunction with additional validation logicimport_filefunction with additional error handling and verification stepsNotes for reviewer:
The changes to the script's code are due to an issue a customer ran into. I was not able to reproduce the issue they ran into (valid file being imported but post-import row-count returns incorrectly then counts again and increments by the amount of rows it counts over and over).
I have focused on potential weaknesses I've identified that could potentially result in such edge-cases, and have refactored and improved their code to be more robust and safer in order to avoid such issues in the future.
I have repeat-tested the script using a dataset that included two large table file parts of each large table, and all small tables in the following two scnearios:
All repeat tests in both scenarios finished successfully.
I have then started a local mirror-importer against the bootstrapped DB and it started cleanly and resumed sync from September 18th timestamps, which is the date this data export was taken on (0.113.2).
Checklist