Re: Backups terminated by client

a_blonde@mindless.com wrote: > > ---- you wrote: > > This is exactly the same problem that I am encountering. Maybe someone can > > shed some light for us. > > I am not very experienced in this area Dudes & Dudettes, but are you looking in the right area? From what I've seen, from version 7.24 onwards, these physical log pages are stored in virtual memory. This memory then grows in segments of size SHMADD. No. The phsysical log pages for pages changed during the archive are stored in temp tables, one per dbspace, in DBSPACETEMP or ROOTDBS as Peter and Peter stated. These temp tables lived until the entire archive was complete in 7.[12]x. In 7.[12]x tables were created non-fragmented in the listed dbspaces round robin so that one of the DBSPACETEMP spaces could fill before the others and crash the archive. > In version 7.30 ontape got 'clever' and actually calculates how much SHM it will need first and then adds one segment of that size. I have seen ontape add a segment of over 250 Mb before! No. In 7.30 the engine got 'clever' and began fragmenting ALL temp tables (including the PHYSLOG temp tables an archive creates) across ALL spaces listed in DBSPACETEMP unless an explicit IN <dbspace> clause is included in the CREATE TEMP TABLE statement. It also began dropping each of the PHYSLOG temp tables after its dbspace is completely archived and that table is dumped to tape. This means that it is FAR less likely that an archive will crash due to not enough temp space since a) less is needed and b) the temp spaces fill roughly evenly. > Is it possible that it is failing to allocate additional SHM, or that SHMTOTAL is set? This process description has been verified by Informix, this is the way it is. Art S. Kagel