Dear all We have two JBODs with 20 or 21 drives available per JBOD hooked up to a server. We are considering the following setups: RAIDZ2 made of 4 drives RAIDZ2 made of 6 drives The first option "wastes" more disk space but can survive a JBOD failure whereas the second is more space effective but the system goes down when a JBOD goes down. Each of the JBOD comes with dual controllers, redundant fans and power supplies so do I need to be paranoid and use option #1? Of course it also gives us more IOPs but high end logging devices should take care of that Thanks for any hint Thomas
On Tue, 31 Jan 2012, Thomas Nau wrote:> Dear all > We have two JBODs with 20 or 21 drives available per JBOD hooked up > to a server. We are considering the following setups: > > RAIDZ2 made of 4 drives > RAIDZ2 made of 6 drives > > The first option "wastes" more disk space but can survive a JBOD failure > whereas the second is more space effective but the system goes down when > a JBOD goes down. Each of the JBOD comes with dual controllers, redundant > fans and power supplies so do I need to be paranoid and use option #1? > Of course it also gives us more IOPs but high end logging devices should take > care of thatI think that the answer depends on the impact to your business if data is temporarily not available. If your business can not survive data being temporarily not available (for hours or even a week) then the more conserative approach may be warranted. If you have a service contract which assures that a service tech will show up quickly with replacement hardware in hand, then this may also influence the decision which should be made. Another consideration is that since these JBODs connect to a server, the data will also be unavailable when the server is down. The server being down may in fact be a more significant factor than a JBOD being down. Bob -- Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
"Hung-Sheng Tsao (Lao Tsao 老曹) Ph.D."
2012-Jan-31 21:05 UTC
[zfs-discuss] need hint on pool setup
what is your main application for ZFS? e.g. just NFS or iSCSI for home dir or VM? or Window client? Is performance important? or space is more important? what is the memory of your server? do you want to use ZIL or L2ARC? what is your backup or DR plan? You need to answer all these question first my 2c On 1/31/2012 3:44 PM, Thomas Nau wrote:> Dear all > We have two JBODs with 20 or 21 drives available per JBOD hooked up > to a server. We are considering the following setups: > > RAIDZ2 made of 4 drives > RAIDZ2 made of 6 drives > > The first option "wastes" more disk space but can survive a JBOD failure > whereas the second is more space effective but the system goes down when > a JBOD goes down. Each of the JBOD comes with dual controllers, redundant > fans and power supplies so do I need to be paranoid and use option #1? > Of course it also gives us more IOPs but high end logging devices should take > care of that > > Thanks for any hint > Thomas > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss-- Hung-Sheng Tsao Ph D. Founder& Principal HopBit GridComputing LLC cell: 9734950840 http://laotsao.blogspot.com/ http://laotsao.wordpress.com/ http://blogs.oracle.com/hstsao/ -------------- next part -------------- A non-text attachment was scrubbed... Name: laotsao.vcf Type: text/x-vcard Size: 608 bytes Desc: not available URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20120131/486b3556/attachment.vcf>
Bob, On 01/31/2012 09:54 PM, Bob Friesenhahn wrote:> On Tue, 31 Jan 2012, Thomas Nau wrote: > >> Dear all >> We have two JBODs with 20 or 21 drives available per JBOD hooked up >> to a server. We are considering the following setups: >> >> RAIDZ2 made of 4 drives >> RAIDZ2 made of 6 drives >> >> The first option "wastes" more disk space but can survive a JBOD failure >> whereas the second is more space effective but the system goes down when >> a JBOD goes down. Each of the JBOD comes with dual controllers, redundant >> fans and power supplies so do I need to be paranoid and use option #1? >> Of course it also gives us more IOPs but high end logging devices should take >> care of that > > I think that the answer depends on the impact to your business if data is temporarily not available. If your business can not > survive data being temporarily not available (for hours or even a week) then the more conserative approach may be warranted.We are talking about home directories at a university so some downtime is ok but fore sure now hours or even days. We do regular backups plus snapshot send-receive to a remote location. The main thing I was wondering about is if it''s better to have a downtime if a JBOD fails (rare I assume) or to keep going without any redundancy left.> If you have a service contract which assures that a service tech will show up quickly with replacement hardware in hand, then > this may also influence the decision which should be made.The replacement hardware is kind of on-site as we use it for the disaster recovery on the remote location> Another consideration is that since these JBODs connect to a server, the data will also be unavailable when the server is down. > The server being down may in fact be a more significant factor than a JBOD being down.I skipped that, sorry. Of course all JOBDs are connected through multiple SAS HBAs to two servers so server failure is easy to handle Thanks for the thoughts Thomas
Hi On 01/31/2012 10:05 PM, "Hung-Sheng Tsao (Lao Tsao ??) Ph.D." wrote:> what is your main application for ZFS? e.g. just NFS or iSCSI for home dir or VM? or Window client?Yes, fileservice only using CIFS, NFS, Samba and maybe iSCSI> Is performance important? or space is more important?a good balance ;)> what is the memory of your server?96G> do you want to use ZIL or L2ARC?ZEUS STECRAM as ZIL (mirrored); maybe SSDs and L2ARC> what is your backup or DR plan?continuous rolling snapshot plus send/receive to remote site TSM backup at least once a week to tape; depends on how much time the TSM client needs to walk the filesystems> You need to answer all these question firstdid so Thomas
my 2c 1 just do mirror of 2 dev with 20 hdd with 1 spare 2 raidz2 with 5 dev for 20 hdd, with one spare Sent from my iPad On Feb 1, 2012, at 3:49, Thomas Nau <Thomas.Nau at uni-ulm.de> wrote:> Hi > > On 01/31/2012 10:05 PM, "Hung-Sheng Tsao (Lao Tsao ??) Ph.D." wrote: >> what is your main application for ZFS? e.g. just NFS or iSCSI for home dir or VM? or Window client? > > Yes, fileservice only using CIFS, NFS, Samba and maybe iSCSI > >> Is performance important? or space is more important? > > a good balance ;) > >> what is the memory of your server? > > 96G > >> do you want to use ZIL or L2ARC? > > ZEUS STECRAM as ZIL (mirrored); maybe SSDs and L2ARC > >> what is your backup or DR plan? > > continuous rolling snapshot plus send/receive to remote site > TSM backup at least once a week to tape; depends on how much > time the TSM client needs to walk the filesystems > >> You need to answer all these question first > > did so > > Thomas >