Re: Ontape Level 0 Archive Aborts Itself....

Jacob Salomon wrote: > > MS CARLEEN C AVANT wrote: > > > > Hi, > > Hoping someone can shed some light on the following. We are running > > Informix ODS 7.24.UC2 on Sun Ultra Enterprise 4000 utilizing Solaris > > 2.6 > > We are using an Exabyte EXB-8500 tape drive; 8mm Helical Scan 7GB > > uncompressed tape. > > > A few days ago we loaded the remainder of our production data. We were > > able to get one successful 0 level archive. It prompted for a second > > tape (for the first time since migration) with no problem. Yesterday > > we received OS Error #28....no space left on device. The Informix > > message log states > > "internal subsystem error...archive aborted by client". > > It never prompted for an additional tape. Here is the key: The first level 0 correctly prompted for tape #2 while the latest ran out of space before prompting! The cause is one of the following: 1) As jake correctly implies if the device is compressed its actual capacity will vary depending on the actual data being written to it, you might try lowering the TAPESIZE as jake suggests. Of course you state that the tape unit is non-compressing! 2) More likely you have run out of DBSPACETEMP space during the archive because of increased user activity. During the archive the engine collects preimages of modified pages from the physical log in temp tables, one per dbspace. Prior to 7.3 temp tables occupied only a single dbspace each and so the different temp tables were created in multiple DBSPACETEMP dbspaces round robin, but there is the possibility that one of these may fill up, especially if the temp tables corresponding to the two busiest dbspaces happen to be written to the same temp dbspace. The temp tables are not released until the archive is completed, even though the preimage pages for each dbspace are written to the tape when that dbspace is completed. You either have to increase the size or number of temp dbspaces, I recommend for the sake of the archives fewer larger dbspaces even though more smaller spaces is better for sorting. Version 7.3 fixes this problem by fragmenting all temp tables across all temp dbspaces. Art S. Kagel