More on Informix Backup Issues

Some of u replied on the note about Informix backup and restore issues and so here's something for u to note: If BLOB data is stored into BLOBSPACEs and during a time when the daily data backup is in progress, AND users try to access the BLOB data in the BLOBSPACE, they will be able to read the BLOB data but they cannot UPDATE or DELETE it... Further more, when the BLOBSPACE is being backed up in the daily data backup, and at the same time IF Informix is performing some HOUSEKEEPING tasks for example a CHECKPOINT, the WHOLE Informix RDMBS hangs for the duration of the BLOB backup INCLUDING being UNABLE to SELECT, UPDATE or DELETE other data (EVEN IF THESE DATA SIT IN DBSPACES !!!) So if your users have stored 300MB of BLOBS in a BLOBSPACE and it takes 45 minutes to backup this BLOBSPACE, and while tbtape is doing a backup of this BLOBSPACE and Informix does a checkpoint (normally at 5 minutes intervals) then your users will be HUNG for about 40 minutes... I spoke to some technical guys from the Informix Secure team in May at Menlo Park... we were actually talking about ODBC and Informix Secure and we jumped onto this BLOB problem... in fact the moment I mentioned it they were real quick to recognise it (guess they must have worked on Informix for quite some time) and I hope they're working on it... Meanwhile if we store BLOB data in DATA SPACEs, we DON'T experience any of the above problems except SLOWER PERFORMANCE for the BLOB data. Slower performance might mean that under high load conditions, eg. if u had 100 users inserting BLOBs at about the same time, your performance might be a few times worse than if they were kept in BLOBSPACES. And if u must use BLOBSPACES and want HIGH AVAILABILITY, then build a daemon process to perform CHECKPOINTS at the desired interval and then when the daemon detects that tbtape is running / or and backing up BLOBSPACES, then suppress the checkpointing. Meanwhile on the tbconfig file, raise checkpoint interval to a very large value... One disadvantage of this approach is that if the system crashes during the backup (and there's a lot of user transactions) during this period then FORWARD RECOVERY will take longer to complete... Cherrio...