Date: Wed, 23 Sep 2015 00:04:19 +0200 From: Willem Jan Withagen <wjw@digiware.nl> To: Mark Felder <feld@FreeBSD.org>, InterNetX - Juergen Gotteswinter <jg@internetx.com>, freebsd-fs@freebsd.org Subject: Re: [HAST] ZFS, many disks, write order Message-ID: <5601D063.4020104@digiware.nl> In-Reply-To: <1442953467.1530841.390840385.067EC1E4@webmail.messagingengine.com> References: <F332F505-F4B3-49D6-8CD7-B53496DCFF3E@gmail.com> <560102BD.6040703@internetx.com> <1442953467.1530841.390840385.067EC1E4@webmail.messagingengine.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 22-9-2015 22:24, Mark Felder wrote: > > > On Tue, Sep 22, 2015, at 02:26, InterNetX - Juergen Gotteswinter wrote: >> One of the great Hast Features is that it Hides Disk Failures from ZFS, >> this has been already discussed on this List a few Months ago. >> >> Single ZIL Flash Drive is also not such a great Idea. I whould suggest >> to start from scratch with "ZFS Best Practices". >> > > A single ZIL is not a single point of failure for a few years now. Most > people do not have a ton of data going through the ZIL and if it dies > you only lose those transactions in flight, not your entire pool. Well if you are talking about HA, then it is not quite in order to lose some transactions. Especially if they originate from synchronous writes that are reported committed to the issuer. So why not have 2 in mirror, either both local, or again in local/remote setup. For caches I'd not use this model, since caches are really not critical. If they die, the data is just re-fetched from disk. --WjW
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5601D063.4020104>