This automatic defragmentation needs an additional parameter to limit the number of containers it will defrag in a day or a max percentage of actual space used that when reached will stop the defrag so it can't fill up the disk to 100% anymore. I see too many cases of Spectrum Protect filling the filesystems to 100% with these processes. It's Saturday and I am working on one right now.
Please look into this IBM. On Fri, Dec 28, 2018 at 1:56 PM Erwann SIMON <erwann.si...@free.fr> wrote: > Hello, > > I've just discovered that the defaults have been changed in 8.1.6 (I've a > test system just upgraded to 8.1.6.100) : > > > Protect : TSMSRV1>q opt defragCntrTrigger > > Server Option Option Setting > ------------------------- ----------------------------------- > defragCntrTrigger 90 > > Protect : TSMSRV1>q opt defragFsTrigger > > Server Option Option Setting > ------------------------- ----------------------------------- > defragFsTrigger 95 > > I was running with 90 and 90 since my last post. > > -- > Best regards / Cordialement / مع تحياتي > Erwann SIMON > > ----- Mail original ----- > De: "Erwann SIMON" <erwann.si...@free.fr> > À: "ADSM: Dist Stor Manager" <ADSM-L@VM.MARIST.EDU> > Envoyé: Jeudi 6 Décembre 2018 17:32:34 > Objet: Re: [ADSM-L] AW: [ADSM-L] Move Container (Automatic) filling up the > directories > > Hello Michael, > > Sorry for the delay and thanks a lot for the advices. I'm trying to find > the right numbers for those two parameters. > > -- > Best regards / Cordialement / مع تحياتي > Erwann SIMON > > ----- Mail original ----- > De: "Michael Prix" <mich...@prix.one> > À: ADSM-L@VM.MARIST.EDU > Envoyé: Jeudi 15 Novembre 2018 16:05:59 > Objet: Re: [ADSM-L] AW: [ADSM-L] Move Container (Automatic) filling up the > directories > > Hello Erwann, > > it depends on what you want to archive. These options were implemented to > reduce the fragmentation of the containers, like the good old reclamation > of > sequential volumes. > Adding to this, it came to the attention of IBM that a container, which > got > moved, should be subject to the reusedelay of the container storage pool. > This > wasn't taken care of until 7.1.8 / 8.1.x. > Speaking of this and putting the pieces together, you may get massive > amount > of "reclamation", the cleaned up containers are subject to reusedelay and > suddenly you are in urgent need for additional and unplanned storage space. > > Speaking of such, setting DEFRAGCNTRTRIGGER and DEFRAGFSTRIFFER to 99 > disables the automatic move container, then you can start lowering both > parameters down to a level which suits you best. > > -- > Michael Prix > > On Thu, 2018-11-15 at 07:06 +0100, Erwann SIMON wrote: > > Hi Michael and Uwe > > > > Thanks *a lot* for your replies. This is excatly what I was looking for. > > > > Did you adjust those options ? How ? > > > > I've was unable to find those server options in the Admin Reference Guide > > and not visible with a standard Q OPT or with a generic Q OPT *DEFR*. I > was > > required to use Q OPT DEFRAGCNTRTRIGGER and Q OPT DEFRAGFSTRIGGER. Both > > options are set to 50% (default). > > > > > > I've found some words in a 8.1.5 presentation : > > > > 8.1.5+ > > > > Defrag uses a new path in MOVE CONTAINER process to read in the chunks > of a > > given container and rewrite them via the standard ingest path, > > redistributing them in available container space > > > > Move Container with defrag allows the user to achieve the goals of > previous > > workarounds much more easily > > Container defrag can be invoked manually, with “MOVE CONTAINER <container > > name> DEFRAG=YES” > > > > Auto-defrag engine periodically checks for container fragmentation > > If containers surpass DEFRAGCNTRTRIGGER & DEFRAGFSTRIGGER thresholds : > > Automatically runs defrags > > If % of available filesystem space or the % of container space which is > > usable/available falls below these thresholds : Auto-defrag will > > periodically start defrag processes for fragmented containers > > > > @Uwe > > For one of the servers (the smallest one, Est. Cap of 12,5TB, Pct. Util > of > > 45%), upgrade has been made at the end of august. The conversion had been > > made monthes ago, when the server was still in 7.1. > > For another one (the biggest, Est. Cap of 235TB, Pct. Util of 65%), > upgrade > > has been made at the end of august too, then I started conversion and it > has > > just finished. > > > > > > -- > > Best regards / Cordialement / مع تحياتي > > Erwann SIMON > > > > ----- Mail original ----- > > De: "Michael Prix" <mich...@prix.one> > > À: ADSM-L@VM.MARIST.EDU > > Envoyé: Mercredi 14 Novembre 2018 19:25:10 > > Objet: Re: [ADSM-L] AW: [ADSM-L] Move Container (Automatic) filling up > the > > directories > > > > Hello, > > > > and, to tell the rest, as both server options are not documented until > > now, > > I quote my documentation from a PMR: > > > > DEFRAGFSTRIGGER 0-99 dynamic server option > > DEFRAGCNTRTRIGGER 0-99 dynamic server option > > > > If > > > > the percentage of free filesystem space is lower than > (100-DEFRAGFSTRIGGER) > > > > and > > > > the percentage of container free space is lower than (100- > > DEFRAGCNTRTRIGGER), > > where those percentages are calculated by > > > > fsfreespace / (fsfreespace + dedup/nondedup cntralloc space) > > > > and > > > > (cntrFree - pool unaccessible ) / ( cntrFree + dedup/nondedup cntrUsed ) > > > > If either threshold is met, then we'll look for containers with a > fraglevel > > higher than 80 to auto-defrag. > > > > >
