# budgeted: cache=True compaction=True
> 1. Deploy r42 to staging failed. What went wrong, according to its log?
call   read_deploy_log({'release': 'r42'})
model  prompt=4502 cached=0 cache=fingerprint only
model  prompt=5261 cached=4401 cache=active
answer The deploy failed during the `pre_deploy` hook while running `alembic upgrade head`. 
> 2. If I roll staging back, does that undo the migration r42 started?
model  prompt=5389 cached=4401 cache=active
answer No. Running `shipctl rollback` moves traffic back to an earlier release but does not run any hooks, so it does not undo database migrations. 
> 3. We fixed the migration and deployed r43, which failed differently. What happened?
call   read_deploy_log({'release': 'r43'})
model  prompt=4792 cached=0 cache=fingerprint only
model  prompt=7664 cached=4691 cache=active
answer The migration succeeded, but the deploy failed its health check. 
> 4. How long are staging logs kept?
model  prompt=7779 cached=4691 cache=active
answer Logs are kept for 14 days in staging (and 30 days in production). Older lines are deleted, so you must export them to your own store if you need them 
> 5. Which of the two failed deploys, r42 or r43, ran its migration to the end?
model  prompt=5735 cached=0 cache=fingerprint only
answer **r43** ran its migration to the end. 
> 6. What is the command to read the full pre_deploy output of r42?
model  prompt=5824 cached=5052 cache=active
answer The command is:
compaction of turns [1, 2]: 1266 characters
compaction of turns [2, 3, 4]: 1302 characters

# plain: cache=False compaction=False
> 1. Deploy r42 to staging failed. What went wrong, according to its log?
call   read_deploy_log({'release': 'r42'})
model  prompt=4502 cached=2469 cache=None
model  prompt=5261 cached=3233 cache=None
answer The deploy failed in the `pre_deploy` hook while running database migrations (`alembic upgrade head`). 
> 2. If I roll staging back, does that undo the migration r42 started?
model  prompt=5397 cached=0 cache=None
answer No. `shipctl rollback` moves traffic to an earlier release but does not run any hooks, so it does not undo database migrations. 
> 3. We fixed the migration and deployed r43, which failed differently. What happened?
call   read_deploy_log({'release': 'r43'})
model  prompt=5490 cached=0 cache=None
model  prompt=8362 cached=0 cache=None
answer Release `r43` failed because of a failed health check. 
> 4. How long are staging logs kept?
model  prompt=8454 cached=0 cache=None
answer Logs are kept for 14 days in staging (and 30 days in production). After that period, older log lines are gone.
> 5. Which of the two failed deploys, r42 or r43, ran its migration to the end?
model  prompt=8508 cached=0 cache=None
answer **r43** ran its migration to the end. 
> 6. What is the command to read the full pre_deploy output of r42?
model  prompt=8601 cached=0 cache=None
answer The command is:
