Date: Fri, 24 Oct 2014 07:52:34 +0100 From: krad <kraduk@gmail.com> To: =?UTF-8?Q?Karli_Sj=C3=B6berg?= <Karli.Sjoberg@slu.se> Cc: "freebsd-fs@freebsd.org" <freebsd-fs@freebsd.org> Subject: Re: How big can ZFS L2ARC grow? Message-ID: <CALfReye-xLGQM=VmA2WckJ8D-5W8Agoz6uNB%2Bc72_iGsVeGtMw@mail.gmail.com> In-Reply-To: <5F9E965F5A80BC468BE5F40576769F099DF6E88A@exchange2-1> References: <5F9E965F5A80BC468BE5F40576769F099DF6E88A@exchange2-1>
next in thread | previous in thread | raw e-mail | index | archive | help
doesn l2arc have compression these days? On 23 October 2014 14:29, Karli Sj=C3=B6berg <Karli.Sjoberg@slu.se> wrote: > Hey! > > As the topic states, I=C2=B4m wondering about the size of L2ARC and if th= ere > is a limit to how big it is able to grow. > > Why I=C2=B4m asking is because I=C2=B4ve always thought that if you add a= cache > device to the pool, the maximum size of L2ARC would be the size of the > disk you=C2=B4ve added, but recently I=C2=B4ve come to know that=C2=B4s n= ot the case. > > Here=C2=B4s a 9.3-RELEASE system that has 64 GB RAM and two 256 GB large > SSD's added as cache, that I would=C2=B4ve thought only could=C2=B4ve gro= wn to > about 512 GB: > # sysctl -n kstat.zfs.misc.arcstats.l2_size > 7696626233344 > > Another system running 9.2-RELEASE with 32 GB RAM + 240 GB SSD: > # sysctl -n kstat.zfs.misc.arcstats.l2_size > 1400038980608 > > The servers are running a software for graphing so I have seen that the > size numbers can go up and down over time, but clearly goes over the > size of the SSD's that have been added. > > We have two more systems configured with cache devices and yet another > two systems configured without. > > The problem we have is that the four systems with cache devices (our > primary storage systems) goes completely unresponsive after different > periods of time, depending on how much RAM they have and the load they > =C2=B4ve been under, I guess. The less RAM, the more frequent they stall,= and > I=C2=B4m starting to wonder if what=C2=B4s common between them is L2ARC, = because > the other two systems without cache devices doesn=C2=B4t have those issue= s, > although they aren=C2=B4t under the same kind of load either, it=C2=B4s a= disaster > recovery system receiving zfs snapshots and the other one is our syslog > server, but still... > > What do you think, are the size numbers for L2ARC unusual, and could it > be related to the stalls we=C2=B4ve been experiencing? And if the size > numbers really are unusual, is there a way to handle it, like limit how > large the L2ARC is able to grow somehow? > > > > -- > > Med V=C3=A4nliga H=C3=A4lsningar > > > -------------------------------------------------------------------------= ------ > Karli Sj=C3=B6berg > Swedish University of Agricultural Sciences Box 7079 (Visiting Address > Kron=C3=A5sv=C3=A4gen 8) > S-750 07 Uppsala, Sweden > Phone: +46-(0)18-67 15 66 > karli.sjoberg@slu.se > _______________________________________________ > freebsd-fs@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-fs > To unsubscribe, send any mail to "freebsd-fs-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CALfReye-xLGQM=VmA2WckJ8D-5W8Agoz6uNB%2Bc72_iGsVeGtMw>