Re: [Linux-HA] Question on order rule
On Thu, Jun 9, 2011 at 4:58 AM, Alessandro Iurlano wrote: > Hello. > > I'm trying to setup an highly available OpenVZ cluster. As OpenVZ only > supports disk quota on ext3/4 local filesystems (nfs and gfs/ocfs2 > don't work), > I have setup two iscsi volumes on an highly available storage where > VMs will be stored. > I would like to use three servers, two active and one spare, for this > cluster. That's because OpenVZ expects all the VMs to be under /vz or > /var/lib/vz directory. This leads to the constraint that each server > can have only on iscsi volume attached and mounted on /vz. > Also, as the ext3 fs is not a cluster filesystem, each iscsi volume > has to be mounted on a single server at time. > > So I need to map two iscsi volumes to three servers. I have created a > pacemaker configuration with two groups, group-iscsi1 and > group-iscsi2, that take care of connecting the iscsi devices and > mounting the filesystems on /vz. A negative colocation directive > forbids the two groups from being active on the same cluster node at > the same time. > > So far things are working. The problem is with the resource that > controls OpenVZ. It is a lsb:vz primitive that needs to be a clone (I > can't make it into separate lsb-vz1 or lsb-vz2 primitives as the > cluster sees both active on every node because they refer to the same > /etc/init.d/vz script). > After creating the clone clone-vz, I defined the location constraints > as location vz-on-iscsi inf: group-iscsi1 clone-vz and did the same > for group-iscsi2. > The clone resource needs to be started after filesystems are mounted, > so I need two order constraints like order vz-after-fs1 inf: > group-iscsi1:start clone-vz:start and order vz-after-fs2 inf: > group-iscsi2:start clone-vz:start > This works perfectly when both iscsi volumes are up. But if one of > them is stopped, clone-vz is not starting. I guess this is because the > two order constraints create dependencies for clone-vz on both the > iscsi groups so that the clone is started only when group-iscsi1 AND > group-iscsi2 have started. > > Is there a way I can tell pacemaker to start clone-vz on a node if > that node has resource group-iscsi1 OR group-iscsi2? Not yet I'm afraid. Resource sets will eventually allow this but I don't think its there yet. > The complete pacemaker configuration is here: > http://nopaste.info/8c2ba79159.html > > Thanks, > Alessandro > ___ > Linux-HA mailing list > Linux-HA@lists.linux-ha.org > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems > ___ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
Re: [Linux-HA] Web resource monitoring
On Mon, Jun 27, 2011 at 3:19 AM, Maxim Ianoglo wrote: > Hello, > >> The http monitoring code should be split off from the apache RA. >> Then a simple stateless (see the Dummy RA for a sample) RA, say >> httpmon, can be created which would source the http monitoring. >> Patches accepted! Guidance and constructive critique offered :) > Ok, thank you for suggestion with name of RA :) > Here is what I wrote: > https://github.com/dotNox/heartbeat_resources/blob/master/httpmon > Small description is available at: > http://dotnox.net/2011/06/multiple-ha-resources-based-on-same-service-heartbeat-httpmon-ra/ > I did not want to patch or change something in apache RA as if someone will > not have any cross of resources, like I have, and it will be easier to use > apache RA. So if apache fails, how does this agent organise recovery? ___ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
[Linux-HA] Help for the floating IPaddress monitoing
Hi Gent, I want to let the floating IP running again after restart the network. But I met the issue when I enable the monitoring for the floating ip (using ocf:heartbeat:IPaddr2). [root@master ~]# crm configure show ip2 primitive ip2 ocf:heartbeat:IPaddr2 \ params ip="172.20.33.88" nic="eth1" iflabel="0" cidr_netmask="255.255.255.0" \ op monitor interval="10s" The IP address configured on the eth0 was lost, and it's so bad that I cannot up the nic before I stop the heartbeat. eth1 Link encap:Ethernet HWaddr 08:00:27:11:87:63 inet6 addr: fe80::a00:27ff:fe11:8763/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:127723 errors:0 dropped:0 overruns:0 frame:0 TX packets:7288 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:11752527 (11.2 MiB) TX bytes:1331914 (1.2 MiB) eth1:0Link encap:Ethernet HWaddr 08:00:27:11:87:63 inet addr:172.20.33.88 Bcast:172.20.33.255 Mask:255.255.255.0 UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 [root@master ~]# ifup eth1 [root@master ~]# ifconfig eth1 eth1 Link encap:Ethernet HWaddr 08:00:27:11:87:63 inet6 addr: fe80::a00:27ff:fe11:8763/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:127956 errors:0 dropped:0 overruns:0 frame:0 TX packets:7362 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:11773863 (11.2 MiB) TX bytes:1341574 (1.2 MiB) I think there maybe a time racing for the '/etc/init.d/network' and 'pacemaker', if 'pacemaker' start the eth1:0 first, then it will not set the IP address for eth1. Does anyone also have the issue? and are there any other way to restart the floating IP without enable the monitoring operation. Regards, -robin ___ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems