When cfops restores the databases and NFS blobstore, it first uploads the backup archives to the /tmp directory. However, the /tmp filesystem only has a capacity of 120MB, which is insufficient for the blobstore and might be for some database backups as well. This causes the restore of the blobstore to fail and renders the platform unusable as the databases are inconsistent with the state of the blobstore, so the cloud controllers fail to start up again. A better alternative for the upload directory might be /var/vcap/store which has a much greater capacity on the database and NFS servers.
aboik
commented
8 years ago
When cfops restores the databases and NFS blobstore, it first uploads the backup archives to the /tmp directory. However, the /tmp filesystem only has a capacity of 120MB, which is insufficient for the blobstore and might be for some database backups as well. This causes the restore of the blobstore to fail and renders the platform unusable as the databases are inconsistent with the state of the blobstore, so the cloud controllers fail to start up again. A better alternative for the upload directory might be /var/vcap/store which has a much greater capacity on the database and NFS servers.