Closed bacongobbler closed 8 years ago
By analyzing the blame information on this pull request, we identified @kmala to be a potential reviewer
I've tested this according to the steps outlined three times, and twice I've deleted the "deis" and all app namespaces and expected helmc install workflow-v2.0.0
to restore everything including my apps, and it's all worked fine.
So LGTM, but I didn't see the original error, so if there are other ways I should test this, let me know.
@mboersma an easy way to test how the error occurs would be to start removing WAL logs from minio:
kd exec -it deis-minio-abd12 bash
cd /home/minio/dbwal/wal_005/
rm -rf *
Then reboot the database. That would expose the checkpoint error.
It appears that the first WAL log is (occasionally) not being shipped to minio, and that's because we did not enable
archive_mode = on
for the first database boot in003_restore_from_backup.sh
, which starts shipping WAL logs immediately after booting. We enable archive mode in 004. This is a simple fix that just requires us enabling archive_mode in003_restore_from_backup.sh
before we boot up the server.closes #114
MANUAL STEPS:
kubectl --namespace=deis delete po deis-database-123ab
)