Re: [openstack-dev] [Manila] Ceph native driver for manila
On 03/04/2015 09:33 AM, Danny Al-Gaaf wrote: Am 04.03.2015 um 15:18 schrieb Csaba Henk: Hi Danny, - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: Deepak Shetty dpkshe...@gmail.com Cc: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Wednesday, March 4, 2015 3:05:46 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... Another level of indirection. I really like the approach of filesystem passthrough ... the only critical question is if virtfs/p9 is still supported in some way (and the question if not: why?). That only seems to be a biggie, isn't it? Yes, it is. We -- Red Hat -- considered a similar, virtfs based driver for GlusterFS but we dropped that plan exactly for virtfs being abandonware. As far as I know it was meant to be a research project, and providing a fairly well working POC it was concluded -- but Deepak knows more of the story. Would like to understand why it was abandoned. I see the need of filesystem passthrough in the area of virtualization. Is there another solution available? Danny, I read through this thread and I wasn't sure I had anything to add, but now that it's gone quiet, I'm wondering what your plan is. I wasn't aware that VirtFS is being considered abandonware but it did seem to me that it wasn't being actively maintained. After looking at the alternatives I considered VirtFS to be the best option for doing what it does, but it's applicability is so narrow that it's hard to find it appealing. I have the following problems with VirtFS: * It requires a QEMU/KVM or Xen hypervisor. VMware and HyperV have zero support nor any plans to support it. * It requires a Linux or BSD. Windows guests can't use it at all. Some googling turned up various projects that might give you a headstart writing a Windows VirtFS client, but we're a long way from having something usable. * VirtFS boils the filesystem down to the bare minimum, thanks to its P9 heritage. Interesting features like caching, locking, security (authentication/authorization/privacy), name mapping, and multipath I/O are either not implemented or delegated to the hypervisor which may or may not meet the needs of the guest application. * Applications designed to run on multiple nodes with shared filesystem storage tend to be tested and supported on NFS or CIFS because those have been around forever. VirtFS is tested and supported by nobody so getting application level support will be impossible. The third one is the one that kills it for me. VirtFS is useful in extremely narrow use cases only. Manila is trying to provide shared filesystems in as wide a set of applications as possible. VirtFS offers nothing that can't also be achieved another way. That's not to say the other way is always ideal. If your use case happens to match exactly what VirtFS does well (QEMU hypervisor, Linux guest, no special filesystem requirements) then the alternatives may not look so good. I'm completely in favor of seeing virtfs support go into Nova and doing integration with it from the Manila side. I'm concerned though that it might be a lot of work, and it might benefits only a few people. Have you found any others who share your goal and are willing to help? Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
On Wed, Mar 4, 2015 at 7:03 AM, Csaba Henk ch...@redhat.com wrote: - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: Csaba Henk ch...@redhat.com, OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org Cc: ceph-de...@vger.kernel.org Sent: Wednesday, March 4, 2015 3:26:52 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila Am 04.03.2015 um 15:12 schrieb Csaba Henk: - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Sunday, March 1, 2015 3:07:36 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use Can you please explain that why does the Ganesha based stack involve 9p? (Maybe I miss something basic, but I don't know.) Sorry, seems that I mixed it up with the p9 case. But the performance is may still an issue if you use NFS on top of CephFS (incl. all the VM layer involved within this setup). For me the question with all these NFS setups is: why should I use NFS on top on CephFS? What is the right to exist of CephFS in this case? I would like to use CephFS directly or via filesystem passthrough. That's a good question. Or indeed, two questions: 1. Why to use NFS? 2. Why does the NFS export of Ceph need to involve CephFS? 1. As of why NFS -- it's probably a good selling point that it's standard filesystem export technology and the tenants can remain backend-unaware as long as the backend provides NFS export. We are working on the Ganesha library -- https://blueprints.launchpad.net/manila/+spec/gateway-mediated-with-ganesha with the aim to make it easy to create Ganesha based drivers. So if you have already an FSAL, you can get at an NFS exporting driver almost for free (with a modest amount of glue code). So you could consider making such a driver for Ceph, to satisfy customers who demand NFS access, even if there is a native driver which gets the limelight. (See commits implementing this under Work Items of the BP -- one is the actual Ganesha library and the other two show how it can be hooked in, by the example of the Gluster driver. At the moment flat network (share-server-less) drivers are supported.) 2. As of why CephFS was the technology chosen for implementing the Ceph FSAL for Ganesha, that's something I'd also like to know. I have the following naive question in mind: Would it not have been better to implement Ceph FSAL with something »closer to« Ceph?, and I have three actual questions about it: - does this question make sense in this form, and if not, how to amend? - I'm asking the question itself, or the amended version of it. - If the answer is yes, is there a chance someone would create an alternative Ceph FSAL on that assumed closer-to-Ceph technology? I don't understand. What closer-to-Ceph technology do you want than native use of the libcephfs library? Are you saying to use raw RADOS to provide storage instead of CephFS? In that case, it doesn't make a lot of sense: CephFS is how you provide a real filesystem in the Ceph ecosystem. I suppose if you wanted to create a lighter-weight pseudo-filesystem you could do so (somebody is building a RadosFS, I think from CERN?) but then it's not interoperable with other stuff. -Greg __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Am 04.03.2015 um 15:18 schrieb Csaba Henk: Hi Danny, - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: Deepak Shetty dpkshe...@gmail.com Cc: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Wednesday, March 4, 2015 3:05:46 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... Another level of indirection. I really like the approach of filesystem passthrough ... the only critical question is if virtfs/p9 is still supported in some way (and the question if not: why?). That only seems to be a biggie, isn't it? Yes, it is. We -- Red Hat -- considered a similar, virtfs based driver for GlusterFS but we dropped that plan exactly for virtfs being abandonware. As far as I know it was meant to be a research project, and providing a fairly well working POC it was concluded -- but Deepak knows more of the story. Would like to understand why it was abandoned. I see the need of filesystem passthrough in the area of virtualization. Is there another solution available? Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Am 04.03.2015 um 15:12 schrieb Csaba Henk: Hi Danny, - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Sunday, March 1, 2015 3:07:36 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use Can you please explain that why does the Ganesha based stack involve 9p? (Maybe I miss something basic, but I don't know.) Sorry, seems that I mixed it up with the p9 case. But the performance is may still an issue if you use NFS on top of CephFS (incl. all the VM layer involved within this setup). For me the question with all these NFS setups is: why should I use NFS on top on CephFS? What is the right to exist of CephFS in this case? I would like to use CephFS directly or via filesystem passthrough. Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Hi Danny, - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: Deepak Shetty dpkshe...@gmail.com Cc: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Wednesday, March 4, 2015 3:05:46 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... Another level of indirection. I really like the approach of filesystem passthrough ... the only critical question is if virtfs/p9 is still supported in some way (and the question if not: why?). That only seems to be a biggie, isn't it? We -- Red Hat -- considered a similar, virtfs based driver for GlusterFS but we dropped that plan exactly for virtfs being abandonware. As far as I know it was meant to be a research project, and providing a fairly well working POC it was concluded -- but Deepak knows more of the story. What's your take on it? I'm really curious if there is any chance for a positive answer. Cheers, Csaba __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Am 04.03.2015 um 05:19 schrieb Deepak Shetty: On Wed, Mar 4, 2015 at 5:10 AM, Danny Al-Gaaf danny.al-g...@bisect.de wrote: Am 03.03.2015 um 19:31 schrieb Deepak Shetty: [...] [...] I was curious to understand. IIUC Neutron provides private and public networks and for VMs to access external CephFS network, the tenant private network needs to be bridged/routed to the external provider network and there are ways neturon achives it. Are you saying that this approach of neutron is insecure ? I don't say neutron itself is insecure. The problem is: we don't want any VM to get access to the ceph public network at all since this would mean access to all MON, OSDs and MDS daemons. If a tenant VM has access to the ceph public net, which is needed to use/mount native cephfs in this VM, one critical issue would be: the client can attack any ceph component via this network. Maybe I misses something, but routing doesn't change this fact. Agree, but there are ways you can restrict the tenant VMs to specific network ports only using neutron security groups and limit what tenant VM can do. On the CephFS side one can use selinux labels to provide addnl level of security for Ceph daemons, where in only certain process can access/modify them, I am just thinking aloud here, i m not sure how well cephfs works with selinux combined. I don't see how neutron security groups would help here. The problem is if a VM has access, in which way ever, to the Ceph network a attacker/user can on one hand attack ALL ceph daemons and on the other also, if there is a bug, crash all daemons and you would lose the complete cluster. SELinux profiles can may help with preventing subvert security or gain privileges it would not help in this case prevent the VM user to crash the cluster. Thinking more, it seems like then you need a solution that goes via the serviceVM approach but provide native CephFS mounts instead of NFS ? Another level of indirection. I really like the approach of filesystem passthrough ... the only critical question is if virtfs/p9 is still supported in some way (and the question if not: why?). Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Hi Danny, - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Sunday, March 1, 2015 3:07:36 PM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila ... For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use Can you please explain that why does the Ganesha based stack involve 9p? (Maybe I miss something basic, but I don't know.) Cheers Csaba __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
On Tue, Mar 3, 2015 at 12:51 AM, Luis Pabon lpa...@redhat.com wrote: What is the status on virtfs? I am not sure if it is being maintained. Does anyone know? The last i knew its not maintained. Also for what its worth, p9 won't work for windows guest (unless there is a p9 driver for windows ?) if that is part of your usecase/scenario ? Last but not the least, p9/virtfs would expose a p9 mount , not a ceph mount to VMs, which means if there are cephfs specific mount options they may not work - Luis - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Sunday, March 1, 2015 9:07:36 AM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila Am 27.02.2015 um 01:04 schrieb Sage Weil: [sorry for ceph-devel double-post, forgot to include openstack-dev] Hi everyone, The online Ceph Developer Summit is next week[1] and among other things we'll be talking about how to support CephFS in Manila. At a high level, there are basically two paths: We discussed the CephFS Manila topic also on the last Manila Midcycle Meetup (Kilo) [1][2] 2) Native CephFS driver As I currently understand it, - The driver will set up CephFS auth credentials so that the guest VM can mount CephFS directly - The guest VM will need access to the Ceph network. That makes this mainly interesting for private clouds and trusted environments. - The guest is responsible for running 'mount -t ceph ...'. - I'm not sure how we provide the auth credential to the user/guest... The auth credentials need to be handled currently by a application orchestration solution I guess. I see currently no solution on the Manila layer level atm. There were some discussion in the past in Manila community on guest auto mount but i guess nothing was conclusive there. Appln orchestration can be achived by having tenant specific VM images with creds pre-loaded or have the creds injected via cloud-init too should work ? If Ceph would provide OpenStack Keystone authentication for rados/cephfs instead of CephX, it could be handled via app orch easily. This would perform better than an NFS gateway, but there are several gaps on the security side that make this unusable currently in an untrusted environment: - The CephFS MDS auth credentials currently are _very_ basic. As in, binary: can this host mount or it cannot. We have the auth cap string parsing in place to restrict to a subdirectory (e.g., this tenant can only mount /tenants/foo), but the MDS does not enforce this yet. [medium project to add that] - The same credential could be used directly via librados to access the data pool directly, regardless of what the MDS has to say about the namespace. There are two ways around this: 1- Give each tenant a separate rados pool. This works today. You'd set a directory policy that puts all files created in that subdirectory in that tenant's pool, then only let the client access those rados pools. 1a- We currently lack an MDS auth capability that restricts which clients get to change that policy. [small project] 2- Extend the MDS file layouts to use the rados namespaces so that users can be separated within the same rados pool. [Medium project] 3- Something fancy with MDS-generated capabilities specifying which rados objects clients get to read. This probably falls in the category of research, although there are some papers we've seen that look promising. [big project] Anyway, this leads to a few questions: - Who is interested in using Manila to attach CephFS to guest VMs? I didn't get this question... Goal of manila is to provision shared FS to VMs so everyone interested in using CephFS would be interested to attach ( 'guess you meant mount?) CephFS to VMs, no ? - What use cases are you interested? - How important is security in your environment? NFS-Ganesha based service VM approach (for network isolation) in Manila is still under works, afaik. As you know we (Deutsche Telekom) are may interested to provide shared filesystems via CephFS to VMs instead of e.g. via NFS. We can provide/discuss use cases at CDS. For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use CephFS directly to the VM would be a bad solution from the security point of view since we can't expose the Ceph public network directly to the VMs to prevent all the security issues we discussed already. Is there any place the security issues are captured for the case where VMs access CephFS directly ? I was curious to understand. IIUC Neutron provides private and public networks and for VMs to access external CephFS network
Re: [openstack-dev] [Manila] Ceph native driver for manila
Am 03.03.2015 um 19:31 schrieb Deepak Shetty: [...] For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use CephFS directly to the VM would be a bad solution from the security point of view since we can't expose the Ceph public network directly to the VMs to prevent all the security issues we discussed already. Is there any place the security issues are captured for the case where VMs access CephFS directly ? No there isn't any place and this is the issue for us. I was curious to understand. IIUC Neutron provides private and public networks and for VMs to access external CephFS network, the tenant private network needs to be bridged/routed to the external provider network and there are ways neturon achives it. Are you saying that this approach of neutron is insecure ? I don't say neutron itself is insecure. The problem is: we don't want any VM to get access to the ceph public network at all since this would mean access to all MON, OSDs and MDS daemons. If a tenant VM has access to the ceph public net, which is needed to use/mount native cephfs in this VM, one critical issue would be: the client can attack any ceph component via this network. Maybe I misses something, but routing doesn't change this fact. Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
On Wed, Mar 4, 2015 at 5:10 AM, Danny Al-Gaaf danny.al-g...@bisect.de wrote: Am 03.03.2015 um 19:31 schrieb Deepak Shetty: [...] For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use CephFS directly to the VM would be a bad solution from the security point of view since we can't expose the Ceph public network directly to the VMs to prevent all the security issues we discussed already. Is there any place the security issues are captured for the case where VMs access CephFS directly ? No there isn't any place and this is the issue for us. I was curious to understand. IIUC Neutron provides private and public networks and for VMs to access external CephFS network, the tenant private network needs to be bridged/routed to the external provider network and there are ways neturon achives it. Are you saying that this approach of neutron is insecure ? I don't say neutron itself is insecure. The problem is: we don't want any VM to get access to the ceph public network at all since this would mean access to all MON, OSDs and MDS daemons. If a tenant VM has access to the ceph public net, which is needed to use/mount native cephfs in this VM, one critical issue would be: the client can attack any ceph component via this network. Maybe I misses something, but routing doesn't change this fact. Agree, but there are ways you can restrict the tenant VMs to specific network ports only using neutron security groups and limit what tenant VM can do. On the CephFS side one can use selinux labels to provide addnl level of security for Ceph daemons, where in only certain process can access/modify them, I am just thinking aloud here, i m not sure how well cephfs works with selinux combined. Thinking more, it seems like then you need a solution that goes via the serviceVM approach but provide native CephFS mounts instead of NFS ? thanx, deepak Danny __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
What is the status on virtfs? I am not sure if it is being maintained. Does anyone know? - Luis - Original Message - From: Danny Al-Gaaf danny.al-g...@bisect.de To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, ceph-de...@vger.kernel.org Sent: Sunday, March 1, 2015 9:07:36 AM Subject: Re: [openstack-dev] [Manila] Ceph native driver for manila Am 27.02.2015 um 01:04 schrieb Sage Weil: [sorry for ceph-devel double-post, forgot to include openstack-dev] Hi everyone, The online Ceph Developer Summit is next week[1] and among other things we'll be talking about how to support CephFS in Manila. At a high level, there are basically two paths: We discussed the CephFS Manila topic also on the last Manila Midcycle Meetup (Kilo) [1][2] 2) Native CephFS driver As I currently understand it, - The driver will set up CephFS auth credentials so that the guest VM can mount CephFS directly - The guest VM will need access to the Ceph network. That makes this mainly interesting for private clouds and trusted environments. - The guest is responsible for running 'mount -t ceph ...'. - I'm not sure how we provide the auth credential to the user/guest... The auth credentials need to be handled currently by a application orchestration solution I guess. I see currently no solution on the Manila layer level atm. If Ceph would provide OpenStack Keystone authentication for rados/cephfs instead of CephX, it could be handled via app orch easily. This would perform better than an NFS gateway, but there are several gaps on the security side that make this unusable currently in an untrusted environment: - The CephFS MDS auth credentials currently are _very_ basic. As in, binary: can this host mount or it cannot. We have the auth cap string parsing in place to restrict to a subdirectory (e.g., this tenant can only mount /tenants/foo), but the MDS does not enforce this yet. [medium project to add that] - The same credential could be used directly via librados to access the data pool directly, regardless of what the MDS has to say about the namespace. There are two ways around this: 1- Give each tenant a separate rados pool. This works today. You'd set a directory policy that puts all files created in that subdirectory in that tenant's pool, then only let the client access those rados pools. 1a- We currently lack an MDS auth capability that restricts which clients get to change that policy. [small project] 2- Extend the MDS file layouts to use the rados namespaces so that users can be separated within the same rados pool. [Medium project] 3- Something fancy with MDS-generated capabilities specifying which rados objects clients get to read. This probably falls in the category of research, although there are some papers we've seen that look promising. [big project] Anyway, this leads to a few questions: - Who is interested in using Manila to attach CephFS to guest VMs? - What use cases are you interested? - How important is security in your environment? As you know we (Deutsche Telekom) are may interested to provide shared filesystems via CephFS to VMs instead of e.g. via NFS. We can provide/discuss use cases at CDS. For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use CephFS directly to the VM would be a bad solution from the security point of view since we can't expose the Ceph public network directly to the VMs to prevent all the security issues we discussed already. We discussed during the Midcycle a third option: Mount CephFS directly on the host system and provide the filesystem to the VMs via p9/virtfs. This need nova integration (I will work on a POC patch for this) to setup libvirt config correctly for virtfs. This solve the security issue and the auth key distribution for the VMs, but it may introduces performance issues due to virtfs usage. We have to check what the specific performance impact will be. Currently this is the preferred solution for our use cases. What's still missing in this solution is user/tenant/subtree separation as in the 2th option. But this is needed anyway for CephFS in general. Danny [1] https://etherpad.openstack.org/p/manila-kilo-midcycle-meetup [2] https://etherpad.openstack.org/p/manila-meetup-winter-2015 -- 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 __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Manila] Ceph native driver for manila
Am 27.02.2015 um 01:04 schrieb Sage Weil: [sorry for ceph-devel double-post, forgot to include openstack-dev] Hi everyone, The online Ceph Developer Summit is next week[1] and among other things we'll be talking about how to support CephFS in Manila. At a high level, there are basically two paths: We discussed the CephFS Manila topic also on the last Manila Midcycle Meetup (Kilo) [1][2] 2) Native CephFS driver As I currently understand it, - The driver will set up CephFS auth credentials so that the guest VM can mount CephFS directly - The guest VM will need access to the Ceph network. That makes this mainly interesting for private clouds and trusted environments. - The guest is responsible for running 'mount -t ceph ...'. - I'm not sure how we provide the auth credential to the user/guest... The auth credentials need to be handled currently by a application orchestration solution I guess. I see currently no solution on the Manila layer level atm. If Ceph would provide OpenStack Keystone authentication for rados/cephfs instead of CephX, it could be handled via app orch easily. This would perform better than an NFS gateway, but there are several gaps on the security side that make this unusable currently in an untrusted environment: - The CephFS MDS auth credentials currently are _very_ basic. As in, binary: can this host mount or it cannot. We have the auth cap string parsing in place to restrict to a subdirectory (e.g., this tenant can only mount /tenants/foo), but the MDS does not enforce this yet. [medium project to add that] - The same credential could be used directly via librados to access the data pool directly, regardless of what the MDS has to say about the namespace. There are two ways around this: 1- Give each tenant a separate rados pool. This works today. You'd set a directory policy that puts all files created in that subdirectory in that tenant's pool, then only let the client access those rados pools. 1a- We currently lack an MDS auth capability that restricts which clients get to change that policy. [small project] 2- Extend the MDS file layouts to use the rados namespaces so that users can be separated within the same rados pool. [Medium project] 3- Something fancy with MDS-generated capabilities specifying which rados objects clients get to read. This probably falls in the category of research, although there are some papers we've seen that look promising. [big project] Anyway, this leads to a few questions: - Who is interested in using Manila to attach CephFS to guest VMs? - What use cases are you interested? - How important is security in your environment? As you know we (Deutsche Telekom) are may interested to provide shared filesystems via CephFS to VMs instead of e.g. via NFS. We can provide/discuss use cases at CDS. For us security is very critical, as the performance is too. The first solution via ganesha is not what we prefer (to use CephFS via p9 and NFS would not perform that well I guess). The second solution, to use CephFS directly to the VM would be a bad solution from the security point of view since we can't expose the Ceph public network directly to the VMs to prevent all the security issues we discussed already. We discussed during the Midcycle a third option: Mount CephFS directly on the host system and provide the filesystem to the VMs via p9/virtfs. This need nova integration (I will work on a POC patch for this) to setup libvirt config correctly for virtfs. This solve the security issue and the auth key distribution for the VMs, but it may introduces performance issues due to virtfs usage. We have to check what the specific performance impact will be. Currently this is the preferred solution for our use cases. What's still missing in this solution is user/tenant/subtree separation as in the 2th option. But this is needed anyway for CephFS in general. Danny [1] https://etherpad.openstack.org/p/manila-kilo-midcycle-meetup [2] https://etherpad.openstack.org/p/manila-meetup-winter-2015 __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
