Re: [zones-discuss] Any issues with migrating zones back-and-forth between sun4u and sun4v systems?
Steve/Steffen, Yes, I just tried this in the lab... - Install base S10U6 on both hosts, create zpool and sparse-root zone. - On T2000, detach zone, and create ZFS snapshot. - zfs send/receive that snapshot to a v490, and import zone with '-u'. - and back to T2000, same way. This works, as long as you have patch 141715-03 (Which requires kernel patch 139555-08), to avoid bug 6802870. However, even though my migrated-back-to-T2000 zone is operating, not sure how stable it is, or what it might be missing. Think I'm going to recommend to the customer that they zfs send their zone ZFS pool to the M5000. Thanks, Jim Steve Lawrence wrote: You can export/import the zpool on shared storage to move the zone. To migrate between sun4v and sun4u, you will need to use attach -u to update the zone to have the necessary platform packages. My guess is that once the zone has been attached to both hosts, you won't need to use the -u after that (assumming niether host is patched). -Steve L. On Fri, Sep 04, 2009 at 11:28:07AM -0400, Steffen Weiberle wrote: On 09/04/09 11:21, Jim Nissen wrote: All, A customer is migrating five zones, from a Niagara system to a M5000, to test performance. The zones are in a ZFS pool, on SAN storage. They plan on moving this storage over to the M5000, and doing a straight import of the zones vs. making a copy of them and then importing them. If the M5000 doesn't perform better, they plan on migrating them back to the Niagara system. I can see that sun4v-to-sun4u zones migrations is supported. However, in the HOW to MOVE A SOLARIS CONTAINER How-To guide, they show an example of using 'zfs send' and 'zfs receive' to make a copy of the zpools. Questions: - Is it supported to just detach the zones and deport the zpool, on one host, and reimport on the other host, without making a separate copy of the zonepath zpool? - How about a straight detach/attach back to the original environment. - If it is supported, has anybody done it? I'm concerned with what might happen to the zone OS bits, if they do a straight import on the M5000, and then go back to the T5220, again. Thanks, Jim ___ zones-discuss mailing list zones-discuss@opensolaris.org I don't have an answer directly, however, I would consider creating a snapshot and clone, and attaching the clone on the other system. then detaching and attaching again on the original system. You could then compare the original with the multi-attached, and look for binary differences. Logs and the like will be different. I do doubt that testing is done in a back and forth manner. Most migrations are one way, or the idea is to have a master and clone from that using the detach/attach methodology. Steffen ___ zones-discuss mailing list zones-discuss@opensolaris.org ___ zones-discuss mailing list zones-discuss@opensolaris.org
Re: [zones-discuss] Will exclusive IP allow for TCP /etc/system settings?
James, No, asking simply because the Solaris tunables guide list both those TCP settings that are changed via ndd and other that are /etc/system. Customer's asking me if exclusive IP will allow for a NGZ to have its own /etc/system TCP tunables, or if they are global. I think Steffen addressed this. Thanks to all, Jim James Carlson wrote: Jim Nissen writes: Will Solaris 10 Zones, with exclusive IP, allow one to set NGZ TCP tunables, like tcp_conn_req_max_q? Yes, every zone configured as exclusive has its own TCP/IP stack instance. Are you asking because you've encountered some problem with this? ___ zones-discuss mailing list zones-discuss@opensolaris.org
[zones-discuss] Will exclusive IP allow for TCP /etc/system settings?
Will Solaris 10 Zones, with exclusive IP, allow one to set NGZ TCP tunables, like tcp_conn_req_max_q? ___ zones-discuss mailing list zones-discuss@opensolaris.org
[zones-discuss] Problem with zones and resource manager pools
We're working with a customer who is using zones, along with Solaris resource manager pools. In short, they have two zones which are assigned to two non-default pools. These two pools each have a processor set, one with one cpu and another with two cpus. Somehow, these two zones started running on the default processor set's processors. One of these zones was rebooted, and that seems to have corrected its processor set mappings. The have a maintenance window to reboot the other zone, but I wanted to do a quick sanity check of any bugs anybody may know of. My sunsolve searches doesn't seem to be turning up any hits on this. Here is a summary of their zone to pool config, with generic names: PoolPset Zone pool_pool1 pset1 (1 cpu)zone1 pool_pool2 pset2 (2 cpus) zone2 pool_defaultpset_default (5 cpus)all other zones In this case, zone1 and zone2 were running on pset_default cpus. Thanks! Jim ___ zones-discuss mailing list zones-discuss@opensolaris.org
Re: [zones-discuss] Problem with zones and resource manager pools
I've been given a possible theory as to how these two zones got into this state (private response). If, for whatever reason, a pool is not available, when a zone comes up/is brought up, the zone will be assigned to the default pool/pset. It is usually accompanied with a pool xxx not available type message. This is highly likely the cause of the state, as customer's zones were in default pool. Plus, when they rebooted their zone, this afternoon, it was assigned to correct pool/pset. Customer did some pool type reconfigs, yesterday, and we've asked them to recall if they got any of these types of messages. Alternatively, the customer could have tried to dynamically bound the zone to a pool/pset, via the 'poolbind -p poolname -i zone zonename' command. Thanks for responses, Jim We're working with a customer who is using zones, along with Solaris resource manager pools. In short, they have two zones which are assigned to two non-default pools. These two pools each have a processor set, one with one cpu and another with two cpus. Somehow, these two zones started running on the default processor set's processors. One of these zones was rebooted, and that seems to have corrected its processor set mappings. The have a maintenance window to reboot the other zone, but I wanted to do a quick sanity check of any bugs anybody may know of. My sunsolve searches doesn't seem to be turning up any hits on this. Here is a summary of their zone to pool config, with generic names: PoolPset Zone pool_pool1 pset1 (1 cpu)zone1 pool_pool2 pset2 (2 cpus) zone2 pool_defaultpset_default (5 cpus)all other zones In this case, zone1 and zone2 were running on pset_default cpus. Thanks! Jim ___ zones-discuss mailing list zones-discuss@opensolaris.org ___ zones-discuss mailing list zones-discuss@opensolaris.org