Why not just move your /var/tmp to the dpool? Seems to me that is a better solution. In any event, I stand by my earlier comments — changing the global default for illumos is not a good idea IMO.
On Wed, Oct 21, 2015 at 9:55 PM, Richard PALO <[email protected]> wrote: > Le 21/10/15 23:47, Garrett D'Amore a écrit : > > I think your problem, Richard, is that your system has put something in > /var/tmp that is badly out of whack with its stated purpose — which is to > provide a “temporary” location on persistent media. Using a USB stick here > is inappropriate for a general purpose system, unless there are no other > options. > > > > If you’re doing a big compile, you *might* want to use /tmp — *IF* you > have enough RAM. Here’s a hint — if that big compile is *illumos* you > better have > 4G RAM there. These days that’s not such a big deal, > although I used to regularly build illumos in VM environments with only 2 > or 4 GB RAM. Which worked fine as long as $TMP was in /var. (Mostly owing > to enormous intermediate files created by lint.) > > I don't believe I have the problem indicated, with 32G ram. > Again, my rpool is an meager Intel SSD, not meant to be intensively > written too > although my dpool is, being a Samsung SSD pro. > > After booting, swap -hs tells me I have ~40GB and: > > richard@omnis:/home/richard$ swap -lh > > swapfile dev swaplo blocks free > > /dev/zvol/dsk/rpool/swap 90,2 4K 2,0G 2,0G > > /dev/zvol/dsk/dpool/swap 90,3 4K 16G 16G > > at the moment, for example: > > richard@omnis:/home/richard$ swap -hs > > total: 1,2G allocated + 306M reserved = 1,5G used, 20G available > > richard@omnis:/home/richard$ echo "::memstat" |pfexec mdb -k > > Page Summary Pages MB %Tot > > ------------ ---------------- ---------------- ---- > > Kernel 2023186 7903 24% > > ZFS File Data 4268560 16674 51% > > Anon 323104 1262 4% > > Exec and libs 3574 13 0% > > Page cache 31462 122 0% > > Free (cachelist) 93174 363 1% > > Free (freelist) 1642014 6414 20% > > > > Total 8385074 32754 > > Physical 8385072 32754 > > Apparently zfs has borrowed 16G from available swap. > > I would *presume* that in any need of swap, space will be returned > but that, I guess, needs to be determined. > I have ~73G avail on dpool... > > What I'm trying to do is preventing rpool from thrashing, using /var/tmp > aggravates that extremely, especially behind my back. > > BTW: > > richard@omnis:/home/richard/src/freebsd/include$ git show 41d91b4c > > commit 41d91b4c7bafe88f7a31de5b7f896a4fd93552cc > > Author: delphij <[email protected]> > > Date: Wed Dec 28 05:35:33 2011 +0000 > > > > In POSIX.1-2008: > > > > P_tmpdir [OB XSI] Default directory prefix for tempnam(). > > > > This macro is used in a lot of places in legacy applications, > > and is why we see a lot of programs written for e.g. Linux > > store volatile temporary files in /var/tmp and not /tmp. > > > > MFC after: 2 months > > > > diff --git a/include/stdio.h b/include/stdio.h > > index 0ae4d8a..02032d9 100644 > > --- a/include/stdio.h > > +++ b/include/stdio.h > > @@ -205,7 +205,7 @@ __END_DECLS > > > > /* System V/ANSI C; this is the wrong way to do this, do *not* use > these. */ > > #if __XSI_VISIBLE > > -#define P_tmpdir "/var/tmp/" > > +#define P_tmpdir "/tmp/" > > #endif > > #define L_tmpnam 1024 /* XXX must be == PATH_MAX */ > > #define TMP_MAX 308915776 > > > -- > Richard PALO > ------------------------------------------- illumos-discuss Archives: https://www.listbox.com/member/archive/182180/=now RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be Modify Your Subscription: https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4 Powered by Listbox: http://www.listbox.com
