DragonFlyBSD

:New submission from Thomas Nikolajsen <thomas.nikolajsen@mail.dk>:
:
:dfly 2.4.1
:Running 'hammer cleanup' stalls fs during 'hammer .. rebalance ..';
:wchan: newbuf (^T):
:disk activity stops & other commands using fs stalls.
:
:Core dump *.42 uploaded to leaf.
:(verbose boot & 'hammer vvv clean' used for this dump)
:
:Info: hammer fs upgraded to v2 after 2.4.0 install,
:didn't use 'hammer rebalance' on fs before 2.4.0.
:'hammer cleanup' run on daily basis on fs;
:didn't see this problem the first few days.
:
:---------
:messages: 7537
:nosy: thomas.nikolajsen
:status: unread
:title: 'hammer rebalance' stalls fs

Thomas, this core seems to be for a system with 128M of ram.
    Is the stall only occurring when the amount of ram is set that low
    or did you set it low to get the crash dump?

We might be hitting up against a base memory requirement for
    running HAMMER, or perhaps some of the limit calculations are
    hitting degenerate cases with such a small buffer cache available.

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>

System only has 128M RAM,
as you have indicated(*) that HAMMER should support this.
It is my local disk server, using an old pc (600MHz),
has worked quite nice until 2.4.

If 128M is too little for a HAMMER+NFS (NAS) now,
I will just use another box for disk server;
just thought system I already had, was an interesting test case.

• e.g. http://leaf.dragonflybsd.org/mailarchive/users/2008-06/msg00075.html

Fixed in commit f6468e9eb2c9 (master) / 12eb5fd70801 (rel2_4)