Integration work
Hi, ceph-devel! It's me, your friendly community guy. Inktank has an engineering team dedicated to Ceph, and we want to work on the right stuff. From time to time, I'd like to check in with you to make sure that we are. Over the past several months, Inktank's engineers have focused on core stability, radosgw, and feature expansion for RBD. At the same time, they have been regularly allocating cycles to integration work. Recently, this has consisted of improvements to the way Ceph works within OpenStack (even though OpenStack isn't the only technology that we think Ceph should play nicely with). What other sorts of integrations would you like to see Inktank engineers work on? For example, are you interested in seeing Inktank spend more of its resources improving interoperability with Apache CloudStack or Eucalyptus? How about Xen? Please share your thoughts. We want to contribute in the best way possible with the resources we have, and your input can help. Thx, Ross -- Ross Turk Community, Ceph @rossturk @inktank @ceph -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Tue, 28 Aug 2012 11:12:16 -0700 Ross Turk wrote: > > Hi, ceph-devel! It's me, your friendly community guy. > > Inktank has an engineering team dedicated to Ceph, and we want to work > on the right stuff. From time to time, I'd like to check in with you to > make sure that we are. > > Over the past several months, Inktank's engineers have focused on core > stability, radosgw, and feature expansion for RBD. At the same time, > they have been regularly allocating cycles to integration work. > Recently, this has consisted of improvements to the way Ceph works > within OpenStack (even though OpenStack isn't the only technology that > we think Ceph should play nicely with). > > What other sorts of integrations would you like to see Inktank engineers > work on? are we only supposed to give answers wrt. integration with other software? if not, I would suggest to write documentation. and also integration with CM like puppet/chef both of these points can give a shorter "time from zero to working cluster" which IMHO is critical in attracting new users. (myself included) -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
Hi Ross, focusing on core stability and feature expansion for RBD was the right appoach in the past and I feel you have reached an adequate maturity level here. Performance enhancements - especially to reduce the latency of a single IO / increase IOPS - and a stronger engagement on the CephFS client would be very much appreciated. A stable and fast CephFS client would allow an efficient integration with - (clustered) NFS (v3 and v4) - (clustered) Samba v4 Cheers, -Dieter On Tue, Aug 28, 2012 at 08:12:16PM +0200, Ross Turk wrote: > > Hi, ceph-devel! It's me, your friendly community guy. > > Inktank has an engineering team dedicated to Ceph, and we want to work > on the right stuff. From time to time, I'd like to check in with you to > make sure that we are. > > Over the past several months, Inktank's engineers have focused on core > stability, radosgw, and feature expansion for RBD. At the same time, > they have been regularly allocating cycles to integration work. > Recently, this has consisted of improvements to the way Ceph works > within OpenStack (even though OpenStack isn't the only technology that > we think Ceph should play nicely with). > > What other sorts of integrations would you like to see Inktank engineers > work on? For example, are you interested in seeing Inktank spend more of > its resources improving interoperability with Apache CloudStack or > Eucalyptus? How about Xen? > > Please share your thoughts. We want to contribute in the best way > possible with the resources we have, and your input can help. > > Thx, > Ross > > -- > Ross Turk > Community, Ceph > @rossturk @inktank @ceph > > > > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
Am 28.08.2012 20:51, schrieb Dieter Kasper: > Performance enhancements - especially to reduce the latency of a single IO / > increase IOPS - > and a stronger engagement on the CephFS client would be very much appreciated. > A stable and fast CephFS client would allow an efficient integration with > - (clustered) NFS (v3 and v4) > - (clustered) Samba v4 Have you tried ocfs2 ontop of rbd in the meanwhile until cephFS gets ready? -- Mit freundlichen Grüßen, Florian Wiessner Smart Weblications GmbH Martinsberger Str. 1 D-95119 Naila fon.: +49 9282 9638 200 fax.: +49 9282 9638 205 24/7: +49 900 144 000 00 - 0,99 EUR/Min* http://www.smart-weblications.de -- Sitz der Gesellschaft: Naila Geschäftsführer: Florian Wiessner HRB-Nr.: HRB 3840 Amtsgericht Hof *aus dem dt. Festnetz, ggf. abweichende Preise aus dem Mobilfunknetz -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Tue, Aug 28, 2012 at 08:57:02PM +0200, Smart Weblications GmbH - Florian Wiessner wrote: > Am 28.08.2012 20:51, schrieb Dieter Kasper: > > > Performance enhancements - especially to reduce the latency of a single IO > > / increase IOPS - > > and a stronger engagement on the CephFS client would be very much > > appreciated. > > A stable and fast CephFS client would allow an efficient integration with > > - (clustered) NFS (v3 and v4) > > - (clustered) Samba v4 > > Have you tried ocfs2 ontop of rbd in the meanwhile until cephFS gets ready? No, I haven't, but I know its limitations. OCFS2 (like GFS/GFS2 from sistina/RH) is build on the cluster-FS design of the 90s. I'm looking for Cluster-FS which is based on + a system which is inherently dynamic + failures in a cluster are the norm, rather than an exception + that characters of workloads are constantly shifting over time + a system which is inevitably built incrementally + a system which is self-managing = Ceph (RBD + CephFS) Cheers, Dieter Kasper > > > -- > > Mit freundlichen Grüßen, > > Florian Wiessner > > Smart Weblications GmbH > Martinsberger Str. 1 > D-95119 Naila > > fon.: +49 9282 9638 200 > fax.: +49 9282 9638 205 > 24/7: +49 900 144 000 00 - 0,99 EUR/Min* > http://www.smart-weblications.de > > -- > Sitz der Gesellschaft: Naila > Geschäftsführer: Florian Wiessner > HRB-Nr.: HRB 3840 Amtsgericht Hof > *aus dem dt. Festnetz, ggf. abweichende Preise aus dem Mobilfunknetz > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- Principal Consultant, Data Center Storage Architecture and Technology FTS CTO FUJITSU TECHNOLOGY SOLUTIONS GMBH Mies-van-der-Rohe-Straße 8 / 4F 80807 München Germany Telephone: +49 89 62060 1898 Telefax:+49 89 62060 329 1898 Mobile: +49 170 8563173 Email: dieter.kas...@ts.fujitsu.com Internet: http://ts.fujitsu.com Company Details: http://ts.fujitsu.com/imprint.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Tue, Aug 28, 2012 at 11:51 AM, Dieter Kasper wrote: > Hi Ross, > > focusing on core stability and feature expansion for RBD was the right appoach > in the past and I feel you have reached an adequate maturity level here. > > Performance enhancements - especially to reduce the latency of a single IO / > increase IOPS - > and a stronger engagement on the CephFS client would be very much appreciated. > A stable and fast CephFS client would allow an efficient integration with > - (clustered) NFS (v3 and v4) > - (clustered) Samba v4 +1 to CephFS being worked on. Things like the multi-mds being improved upon would be amazing. Regards, Tren -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 08/28/2012 11:32 AM, Plaetinck, Dieter wrote: > On Tue, 28 Aug 2012 11:12:16 -0700 > Ross Turk wrote: > >> >> Hi, ceph-devel! It's me, your friendly community guy. >> >> Inktank has an engineering team dedicated to Ceph, and we want to work >> on the right stuff. From time to time, I'd like to check in with you to >> make sure that we are. >> >> Over the past several months, Inktank's engineers have focused on core >> stability, radosgw, and feature expansion for RBD. At the same time, >> they have been regularly allocating cycles to integration work. >> Recently, this has consisted of improvements to the way Ceph works >> within OpenStack (even though OpenStack isn't the only technology that >> we think Ceph should play nicely with). >> >> What other sorts of integrations would you like to see Inktank engineers >> work on? > > are we only supposed to give answers wrt. integration with other software? > if not, I would suggest to write documentation. If I may say so, the amount of work that John has poured into this in recent week has been incredible (http://www.ceph.com/docs/master/). So while it's definitely not complete nor perfect, I'm sure he would appreciate a little more specific information as to where you believe documentation is lacking. I for my part, in the documentation space, would love for the admin tools to become self-documenting. For example, I would love a "help" subcommand at any level of the ceph shell, listing the supported subcommands in that level. As in "ceph help", "ceph mon help", "ceph osd getmap help". Even better, the ceph shell could support a general-purpose hook that bash-completion can use (kind of like "hg" does in Mercurial), and this and the above-conjectured help facility could arguably share quite a bit of code. > and also integration with CM like puppet/chef +1, although people are already working on both. So maybe this is just about the need to tell more people about that. :) Cheers, Florian -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Tue, Aug 28, 2012 at 5:03 PM, Florian Haas wrote: > I for my part, in the documentation space, would love for the admin > tools to become self-documenting. For example, I would love a "help" > subcommand at any level of the ceph shell, listing the supported > subcommands in that level. As in "ceph help", "ceph mon help", "ceph osd > getmap help". > > Even better, the ceph shell could support a general-purpose hook that > bash-completion can use (kind of like "hg" does in Mercurial), and this > and the above-conjectured help facility could arguably share quite a bit > of code. I would love to see all of that. But, a lot of the "ceph" tool functionality is implemented by shoveling strings in and out of the monitors. It largely doesn't understand what's happening. If we were to redo that from scratch, I'd convert that to have some sort of API to monitors, and make the cli understand all the relevant things. Understandably, that can feel a little bit more rigid; to add a command means adding it to both the server and a client, where as currently the client is very very generic. >> and also integration with CM like puppet/chef > +1, although people are already working on both. So maybe this is just > about the need to tell more people about that. :) Please do give constructive feedback on http://ceph.com/docs/master/install/chef/ http://ceph.com/docs/master/config-cluster/chef/ -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 08/28/2012 02:15 PM, Tommi Virtanen wrote: On Tue, Aug 28, 2012 at 5:03 PM, Florian Haas wrote: I for my part, in the documentation space, would love for the admin tools to become self-documenting. For example, I would love a "help" subcommand at any level of the ceph shell, listing the supported subcommands in that level. As in "ceph help", "ceph mon help", "ceph osd getmap help". Even better, the ceph shell could support a general-purpose hook that bash-completion can use (kind of like "hg" does in Mercurial), and this and the above-conjectured help facility could arguably share quite a bit of code. I would love to see all of that. But, a lot of the "ceph" tool functionality is implemented by shoveling strings in and out of the monitors. It largely doesn't understand what's happening. It doesn't need to understand what's happening to give basic usage info though - the monitors can provide that themselves in the short term while we don't have an admin api like you describe below. I added a feature request for this a little while back: http://www.tracker.newdream.net/issues/2894 If we were to redo that from scratch, I'd convert that to have some sort of API to monitors, and make the cli understand all the relevant things. Understandably, that can feel a little bit more rigid; to add a command means adding it to both the server and a client, where as currently the client is very very generic. -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Tuesday 28 August 2012 you wrote: > On Tue, Aug 28, 2012 at 11:51 AM, Dieter Kasper wrote: > > Hi Ross, > > > > focusing on core stability and feature expansion for RBD was the right > > appoach in the past and I feel you have reached an adequate maturity > > level here. > > > > Performance enhancements - especially to reduce the latency of a single > > IO / increase IOPS - and a stronger engagement on the CephFS client would > > be very much appreciated. A stable and fast CephFS client would allow an > > efficient integration with - (clustered) NFS (v3 and v4) > > - (clustered) Samba v4 > > +1 to CephFS being worked on. Things like the multi-mds being improved > upon would be amazing. Stable CephFS is what we need, too - many concurrent write accesses from multiple clients with a real file system underneath. However, most stability problems we have had so far were crashes in the daemons, not the Linux kernel. Great for me would be some solution to what we call the Domino effect - one daemon crashes, the next takes over, crashes at the same place (same data...), until the whole cluster is down. There will always be bugs, but they should not kill the whole cluster. In our tests, single MDS was no real bottleneck, it was only lacking stability. I have not tested the newest releases, so it might be better now. Improved performance with many small files being written concurrently would be great, but CephFS has been getting significantly faster over the last year and performance is being worked on all the time. Amon Ott -- Dr. Amon Ott m-privacy GmbH Tel: +49 30 24342334 Am Köllnischen Park 1Fax: +49 30 99296856 10179 Berlin http://www.m-privacy.de Amtsgericht Charlottenburg, HRB 84946 Geschäftsführer: Dipl.-Kfm. Holger Maczkowsky, Roman Maczkowsky GnuPG-Key-ID: 0x2DD3A649 -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
Hi, > How about Xen? I vote for this :) Using RBD storage for Xen VM images / disks is IMHO a very nice fit, the same way people do with QEMU. This should even allow live migration of VM. Currently we have to rely on the RBD kernel driver which has some downsides (no caching / need recent kernel to get latest ceph patches). There also seem to be some weird interactions between RBD and Xen that lead to significant performance hits that are not present when using only RBD or only Xen. One possibility would be to develop a blktap driver for xen to provide block device backend in userspace using librbd rather than kernel mode rbd. Cheers, Sylvain -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 08/29/2012 10:20 AM, Sylvain Munaut wrote: Hi, How about Xen? I vote for this :) Using RBD storage for Xen VM images / disks is IMHO a very nice fit, the same way people do with QEMU. This should even allow live migration of VM. Correct me if I'm wrong, but when I was at Citrix in May this year somebody there told me that Xen was going 100% Qemu? By going 100% Qemu they would also get RBD support. Wido Currently we have to rely on the RBD kernel driver which has some downsides (no caching / need recent kernel to get latest ceph patches). There also seem to be some weird interactions between RBD and Xen that lead to significant performance hits that are not present when using only RBD or only Xen. One possibility would be to develop a blktap driver for xen to provide block device backend in userspace using librbd rather than kernel mode rbd. Cheers, Sylvain -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
> Correct me if I'm wrong, but when I was at Citrix in May this year somebody > there told me that Xen was going 100% Qemu? Huh ... I've never heard this. Also the guys in ##xen haven't either. I'm not really involved in xen dev and don't follow it closely but that seems unlikely. The few slides I looked at from the Xen Summit a couple days ago show that they really like their PV model. AFAIK QEMU is only used for HVM guests to emulate the hw. And even for those HVM guest, it's recommended to use PV drivers for performance which bypass the qemu layer all together. Cheers, Sylvain -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 08/29/2012 02:35 PM, Sylvain Munaut wrote: Correct me if I'm wrong, but when I was at Citrix in May this year somebody there told me that Xen was going 100% Qemu? Huh ... I've never heard this. Also the guys in ##xen haven't either. I'm not really involved in xen dev and don't follow it closely but that seems unlikely. The few slides I looked at from the Xen Summit a couple days ago show that they really like their PV model. I must be wrong then! AFAIK QEMU is only used for HVM guests to emulate the hw. And even for those HVM guest, it's recommended to use PV drivers for performance which bypass the qemu layer all together. Not sure where this came from then, but in that case it would take work from Citrix to get RBD in Xen. Wido Cheers, Sylvain -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Wed, Aug 29, 2012 at 9:40 AM, Wido den Hollander wrote: >> Huh ... I've never heard this. Also the guys in ##xen haven't either. >> I'm not really involved in xen dev and don't follow it closely but >> that seems unlikely. The few slides I looked at from the Xen Summit a >> couple days ago show that they really like their PV model. > I must be wrong then! They are (at least, Red Hat is) looking at using more qemu for xen-hvm. Whether that has any effect on the PV side, I wouldn't know. It might make sense for them to use virtio even for PV, so they might use qemu to implement the hypervisor side of virtio too, and that would get you librbd support. -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 29 August 2012 23:43, Tommi Virtanen wrote: > On Wed, Aug 29, 2012 at 9:40 AM, Wido den Hollander wrote: >>> Huh ... I've never heard this. Also the guys in ##xen haven't either. >>> I'm not really involved in xen dev and don't follow it closely but >>> that seems unlikely. The few slides I looked at from the Xen Summit a >>> couple days ago show that they really like their PV model. >> I must be wrong then! > > They are (at least, Red Hat is) looking at using more qemu for > xen-hvm. Whether that has any effect on the PV side, I wouldn't know. > It might make sense for them to use virtio even for PV, so they might > use qemu to implement the hypervisor side of virtio too, and that > would get you librbd support. > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html I don't there is much going on in terms of increasing use of QEMU, only that Xen can now use upsteam QEMU rather than the Xen specific fork (qemu-xen-traditional). There was GSOC project to build a virtio front/backend for Xen but I am not sure if this would be the way to go. As far as I can see Xen dominates KVM in terms network and I/O performance on every benchmark so apart from compatibility the gains of using virtio don't seem that great... Xen's blkback/netback PV system is just that much faster and more scalable with large numbers of domains or 100k+ IOPs. With regards to blktap.. blktap is currently in a state where blktap2 is included in a minimal amount of distros and is non-upstreamable. blktap3 which is coming will but fully userspace but I have never been a big fan of userspace block devices, YMMV. That being said, building blktap devices is really easy (similar to tuntap). Ideally improving the kernel RBD device would provide the best performance across the board and the most compatibility (anything can use a raw block device). Joseph. -- CTO | Orion Virtualisation Solutions | www.orionvm.com.au Phone: 1300 56 99 52 | Mobile: 0428 754 846 -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On 08/28/2012 10:20 PM, Josh Durgin wrote: > On 08/28/2012 02:15 PM, Tommi Virtanen wrote: >> On Tue, Aug 28, 2012 at 5:03 PM, Florian Haas >> wrote: >>> I for my part, in the documentation space, would love for the admin >>> tools to become self-documenting. For example, I would love a "help" >>> subcommand at any level of the ceph shell, listing the supported >>> subcommands in that level. As in "ceph help", "ceph mon help", "ceph osd >>> getmap help". >>> >>> Even better, the ceph shell could support a general-purpose hook that >>> bash-completion can use (kind of like "hg" does in Mercurial), and this >>> and the above-conjectured help facility could arguably share quite a bit >>> of code. >> >> I would love to see all of that. But, a lot of the "ceph" tool >> functionality is implemented by shoveling strings in and out of the >> monitors. It largely doesn't understand what's happening. > > It doesn't need to understand what's happening to give basic usage info > though - the monitors can provide that themselves in the short term > while we don't have an admin api like you describe below. > > I added a feature request for this a little while back: > > http://www.tracker.newdream.net/issues/2894 I believe this is pretty straightforward to get done. -- João Eduardo Luís gpg key: 477C26E5 from pool.keyserver.eu signature.asc Description: OpenPGP digital signature
RE: Integration work
Ross, All: I've read through several recommendations, and I'd like to add 2 to that list for consideration. First: For my local project, I'm using rbd with Oracle VM and VM manager, mainly because of the other engineers' familiarity with the Oracle platforms, and they're certified by MS to run Windows on Xen (using Oracle's stuff). Now, due to necessity, I'll be working on a Storage plugin that allows the Orcale VM to understand RBD, to make pools, etc for our project. I would be interested to know if anyone else has actually started on their own version of the same. Secondly: Through some trials, I've found that if one loses all of his Monitors in a way that they also lose their disks, one basically loses their cluster. I would like to recommend a lower priority shift in design that allows for "recovery of the entire monitor set from data/snapshots automatically stored at the osd's". For example, a monitor boots: -keyring file and ceph.conf are available -monitor sees that it is missing its local copy of maps, etc. -goes onto the first OSD's it sees and pulls down a snapshot of the same -checks for another running monitor, syncs with it, if not, -boots at quorum 0, verifying OSD states -life continues. The big deal here, is that while the entire cluster is able to recover from failures using one storage philosophy, the monitors are using an entirely different, and more legacy storage philosophy - basically local RAID/power in numbers. Perhaps this has already been considered, and I would be interested in knowing what people think here, as well. Or perhaps I missed something and this is already done? Thanks, for your time! Ryan Nicholson -Original Message- From: ceph-devel-ow...@vger.kernel.org [mailto:ceph-devel-ow...@vger.kernel.org] On Behalf Of Ross Turk Sent: Tuesday, August 28, 2012 1:12 PM To: ceph-devel@vger.kernel.org Subject: Integration work Hi, ceph-devel! It's me, your friendly community guy. Inktank has an engineering team dedicated to Ceph, and we want to work on the right stuff. From time to time, I'd like to check in with you to make sure that we are. Over the past several months, Inktank's engineers have focused on core stability, radosgw, and feature expansion for RBD. At the same time, they have been regularly allocating cycles to integration work. Recently, this has consisted of improvements to the way Ceph works within OpenStack (even though OpenStack isn't the only technology that we think Ceph should play nicely with). What other sorts of integrations would you like to see Inktank engineers work on? For example, are you interested in seeing Inktank spend more of its resources improving interoperability with Apache CloudStack or Eucalyptus? How about Xen? Please share your thoughts. We want to contribute in the best way possible with the resources we have, and your input can help. Thx, Ross -- Ross Turk Community, Ceph @rossturk @inktank @ceph -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Integration work
On Fri, Aug 31, 2012 at 11:02 PM, Ryan Nicholson wrote: > Secondly: Through some trials, I've found that if one loses all of his > Monitors in a way that they also lose their disks, one basically loses their > cluster. I would like to recommend a lower priority shift in design that > allows for "recovery of the entire monitor set from data/snapshots > automatically stored at the osd's". > > For example, a monitor boots: > -keyring file and ceph.conf are available > -monitor sees that it is missing its local copy of maps, etc. > -goes onto the first OSD's it sees and pulls down a snapshot of the > same > -checks for another running monitor, syncs with it, if not, > -boots at quorum 0, verifying OSD states > -life continues. Monitor fetching initial information from an OSD is full of challenges. The monitor won't know what IP addresses and ports the OSDs are, the OSDs won't trust the monitor to talk to them, etc (it lost its crypto keys, after all). It wouldn't even know which OSD to talk to, and I highly doubt having the backup on every OSD would be a good idea. > The big deal here, is that while the entire cluster is able to recover from > failures using one storage philosophy, the monitors are using an entirely > different, and more legacy storage philosophy - basically local RAID/power in > numbers. Perhaps this has already been considered, and I would be interested > in knowing what people think here, as well. Or perhaps I missed something and > this is already done? That's why you run multiple monitors: they provide High Availability to the monitor service, as a whole. Losing all of your monitors at all disrupts operation of the cluster. Losing all of their stable storage really is disastrous. This is why you are supposed to deploy them in different failure domains, e.g. in different rows or rooms. If a monitor has its mon. keyring and ceph.conf, it should be able to join an existing monitor cluster as a new member, no special-case recovery needed. I'm not sure what kind of architecture you have that makes losing all the of the monitor disks somehow likely, but perhaps you should just take backups of their disks, with plain-old backup tools? Don't try to store that backup in the same Ceph cluster, though. It would be interesting to hear more about what you're thinking of, here. -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html