Common Properties¶
All service resources share common properties and behaviours.
Action Requirements¶
to be completed
Tagging¶
A resource can be tagged using the keyword tags. The value is a whitespace-separated list of tags. Tag names can be user-defined or hardcoded in the agent.
Custom Tags¶
Custom tags ease service management on complex configurations, as they can be used in services and resources selector expressions.
Examples:
| Resource | Tags |
|---|---|
| app#db | database base |
| app#tomcat1 | appsrv base |
| app#tomcat2 | appsrv |
| app#tomcat3 | appsrv |
| app#nginx1 | websrv base |
| app#nginx2 | websrv |
# stop resources tagged 'websrv'
$ om <svcname> --tag websrv stop
# stop resources tagged 'websrv' or 'appsrv'
$ om <svcname> --tag websrv,appsrv stop
# stop resources tagged 'websrv' and 'base'
$ om <svcname> --tag websrv+base stop
Special Tags¶
Some tag names are reserved and have a particular meaning.
noaction¶
This tag keep the agent from executing state-changing actions on the resource.
The agent is still running the resource status evaluations.
For example, the resource mapping the ip address activated at vm boot by the operating system must be tagged noaction.
encap¶
This tag assigns the resource to the encapsulated/slave service. The agent on the master-part of the service does not handle such a resource.
svcmgr print status highlights such resources with the E flag.
$ om testzone print status testzone up `- sol3.opensvc.com up frozen, | idle, | started |- disk#0 ..../ n/a testzone.raw0 |- disk#1 ..../ up testzone.raw1 |- fs#0 ..... n/a dir /tmp/share |- share#1 ..../ up nfs:/tmp/share `- container#0 ..../ up testzone |- ip#1 ...E/ down 128.0.1.2@lo0/testzone1 `- app:a1 // |- app#0 ...E/ n/a true | info: not evaluated | (instance not up) `- app#1 ...E/ n/a true info: not evaluated (instance not up)
See also
nostatus¶
This tag prevents the resource status evaluation. The resource status is set to n/a.
dedicated¶
This tag is by the ip.docker driver only. If set, the physical network interface card is moved to the container network namespace. This NIC is thus reserved, and should not be used by other resources and services.
Subsets¶
to be completed
Disabled¶
A resource can be marked as disabled using the disable keyword.
[container#1] type = docker image = ubuntu:14.04 interactive = true tty = true entrypoint = /bin/bash disable = true
This will make the agent ignore any action upon this resource.
svcmgr print status will highlights disabled resources with the D flag.
$ om app1.dev print status --refresh app1.dev up `- deb1.opensvc.com up idle, started |- ip#0 ..... up 192.168.1.1@lo `- container#1 .D... n/a docker container app1.dev.container.1@ubuntu:14.04
Optional¶
A resource can be marked as optional using the optional keyword.
[app#0] script = /bin/true info = true stop = true start = true optional = true
This parameter allow defining non critical resources in the service.
Service actions won't stop on error reported by optional resources.
svcmgr print status will highlights optional resources with the O flag.
$ om redis.acme.com print status mysvc1.opensvc.com up `- deb1.opensvc.com up idle, started |- ip#1 ..... up 128.0.1.124@lo |- disk#1 ..... stdby up loop /opt/disk1.dd |- disk#2 ..... stdby up loop /opt/disk2.dd |- disk#3 ..... stdby up vg vgtest |- fs#1 ..... up ext4 /dev/vgtest/lvtest1@/opt/avn/lvtest1 |- fs#2 ..... up ext4 /dev/vgtest/lvtest2@/opt/avn/lvtest2 |- fs#3 ..... up ext4 /dev/disk/by-label/testfs@/opt/avn/lvtest3 |- share#0 ..../ up nfs:/opt/avn/lvtest3 |- app#0 ..O./ n/a true | info: check is not set `- sync#i0 ..O./ up rsync svc config to drpnodes, nodes
Monitoring¶
A resource can be marked as monitored using the monitor keyword.
[disk#3] type = vg name = vgtest standby = true monitor = true
It means that this resource is critical for the service availability.
If the resource goes down, then the agent triggers the monitor_action, which may cause a crash or reboot of the node, or stop of the service, to force a failover.
svcmgr print status will highlights monitored resources with the M flag.
$ om redis.acme.com print status mysvc1.opensvc.com up `- deb1.opensvc.com up idle, started |- ip#1 ..... up 128.0.1.124@lo |- disk#1 ..... stdby up loop /opt/disk1.dd |- disk#2 ..... stdby up loop /opt/disk2.dd |- disk#3 M.... stdby up vg vgtest |- fs#1 ..... up ext4 /dev/vgtest/lvtest1@/opt/avn/lvtest1 |- fs#2 ..... up ext4 /dev/vgtest/lvtest2@/opt/avn/lvtest2 |- fs#3 ..... up ext4 /dev/disk/by-label/testfs@/opt/avn/lvtest3 |- share#0 ..../ up nfs:/opt/avn/lvtest3 |- app#0 ..O./ n/a true | info: check is not set `- sync#i0 ..O./ up rsync svc config to drpnodes, nodes
Note
restart parameter can be combined with monitor setting, as explained below
Automatic Restart¶
The restart parameter can be set to make the agent daemon monitor restart the resource if it fails:
[app#0] script = /bin/true info = true stop = true start = true optional = true restart = 2
The restart value is the number of times the daemon will attempt to restart the resource before giving up.
If combined with monitor, the agent will try to restart the failed resource before triggering the monitor_action
Standby resources¶
Some resources must remain up, even when the service instance is stopped.
For example, in a 2-nodes failover service with a fs resource and a sync.rsync resource replicating the fs, the fs resource must be up on the passive node receive the rsync'ed data. If not, the data gets written to the underlying filesystem.
The standby keyword can be set in these cases:
[disk#3] type = vg name = vgtest standby = true monitor = true
Possible values are 'nodes', 'drpnodes' or 'nodes drpnodes', or a list of nodes.
Resources with standby = true are started on service boot and start actions, and stopped only on service shutdown action.
svcgr print status will display the stdby up status for up standby resources, and stdby down status for down standby resources.
# Primary Node $ om mysvc.acme.com print status mysvc1.opensvc.com up `- deb1.opensvc.com up idle, started |- ip#1 ..... up 128.0.1.124@lo |- disk#1 ..... stdby up loop /opt/disk1.dd |- disk#2 ..... stdby up loop /opt/disk2.dd |- disk#3 M.... stdby up vg vgtest |- fs#1 ..... up ext4 /dev/vgtest/lvtest1@/opt/avn/lvtest1 |- fs#2 ..... up ext4 /dev/vgtest/lvtest2@/opt/avn/lvtest2 |- fs#3 ..... up ext4 /dev/disk/by-label/testfs@/opt/avn/lvtest3 |- share#0 ..../ up nfs:/opt/avn/lvtest3 |- app#0 ..O./ n/a true | info: check is not set `- sync#i0 ..O./ up rsync svc config to drpnodes, nodes # Secondary Node mysvc1.opensvc.com `- deb2.opensvc.com warn warn |- ip#1 ..... down 128.0.1.124@lo |- disk#1 ..... stdby up loop /opt/disk1.dd |- disk#2 ..... stdby down loop /opt/disk2.dd |- disk#3 M.... stdby up vg vgtest |- fs#1 ..... down ext4 /dev/vgtest/lvtest1@/opt/avn/lvtest1 |- fs#2 ..... down ext4 /dev/vgtest/lvtest2@/opt/avn/lvtest2 |- fs#3 ..... down ext4 /dev/disk/by-label/testfs@/opt/avn/lvtest3 |- share#0 ..../ down nfs:/opt/avn/lvtest3 |- app#0 ..O.. n/a true | info: not evaluated (instance not up) `- sync#i0 ..O./ up rsync svc config to drpnodes, nodes
Warning
Don't set shared disk always on. This would cause data corruption.
Triggers¶
Triggers are scripts or commands executed before or after a resource action.
Syntax¶
In a service resource configuration section, triggers can be defined using a keyword formatted as
[blocking_]<when>_<action>[@<scope>] = <command>
Blocking¶
If the blocking_ prefix is set, a non-zero command return code causes the service action to abort on error.
If the blocking_ prefix is not set, a non-zero command return code causes an error log, but the service action continues.
When¶
Supported values:
preThe command is executed before the normal resource driver action.postThe command is executed after the normal resource driver action.
Action¶
Supported values:
startstoprunprovisionunprovisionsync_nodessync_drpsync_resyncsync_update
Command¶
The command value supports shell expressions, like &&, ||, ...
Warning
; is interpreted as a comment begin marker if preceded by a whitespace. To use this character as a separator between commands, be sure to not insert a whitespace before. Example: post_start = echo hey; echo ya is valid but post_start = echo hey ; echo ya is not.
Use Cases¶
Inter-service dependencies
Using a
pre_starttrigger to implement a wait loop for a tiers service expected state. Ablocking_pre_starttrigger can be combined with a wait loop timeout to abort the service startup if the required tiers service does not reach the expected state in time.Service provisioning
Using a
post_provisiontrigger to create directories, files, run configuration management, deploy app, ... just after the filesystems are provisioned and started.Extra start checks
Using a
blocking_pre_starton the first started resource to verify extra conditions to allow service instance startup.
Resource Devices¶
The agent reports to the collector the base disks list and and size. For each disk, it also reports the size used by each service.
$ om node pushdisks aubergine Size.Used Vendor Model `- disks |- 002538b471bb6f3c 953g SAMSUNG MZSLW1T0HMLH-000L1 | |- testdrbd 3g | |- testmd 991m | |- pridns 6g | |- testmd2 143m | `- aubergine 943g |- aubergine.md125 0 |- aubergine.md127 0 `- aubergine.md126 0
This feature depends on a device tree build by the agent.
This tree can be display bottom-up:
aubergine Type Size Pct of Parent |- loop1 linear 0 - |- md127 linear 0 - |- md125 linear 0 - |- loop6 linear 0 - |- loop4 linear 0 - |- loop2 linear 0 - |- md126 linear 0 - |- loop7 linear 0 - |- nvme0n1 linear 953g - | |- nvme0n1p1 linear 512m 0% | `- nvme0n1p2 linear 953g 99% | |- ubuntu--vg-swap_1 linear 15g 1% | `- ubuntu--vg-root linear 915g 96% | |- loop13 linear 10m 0% | |- loop12 linear 50m 0% | |- loop11 linear 50m 0% | | `- md122 raid1 49m 98% | | `- md123 raid0 97m 197% | |- loop10 linear 50m 0% | | `- md122 raid1 49m 98% | | `- md123 raid0 97m 197% | |- loop14 linear 143m 0% | |- loop9 linear 50m 0% | | `- md124 raid1 49m 98% | | `- md123 raid0 97m 197% | |- loop8 linear 50m 0% | | `- md124 raid1 49m 98% | | `- md123 raid0 97m 197% | `- loop0 linear 3g 0% | `- drbd1 0 0% |- loop5 linear 0 - `- loop3 linear 0 -
Or top-bottom:
$ om node print devs --reverse aubergine Type Parent Use Size Ratio |- drbd1 - 0 - | `- loop0 linear 0 3g - | `- ubuntu--vg-root linear 3g 915g 0% | `- nvme0n1p2 linear 915g 953g 96% | `- nvme0n1 linear 953g 953g 99% |- md123 raid0 - 97m - | |- md124 raid1 48m 49m 97% | | |- loop8 linear 49m 50m 98% | | | `- ubuntu--vg-root linear 50m 915g 0% | | | `- nvme0n1p2 linear 915g 953g 96% | | | `- nvme0n1 linear 953g 953g 99% | | `- loop9 linear 49m 50m 98% | | `- ubuntu--vg-root linear 50m 915g 0% | | `- nvme0n1p2 linear 915g 953g 96% | | `- nvme0n1 linear 953g 953g 99% | `- md122 raid1 48m 49m 97% | |- loop11 linear 49m 50m 98% | | `- ubuntu--vg-root linear 50m 915g 0% | | `- nvme0n1p2 linear 915g 953g 96% | | `- nvme0n1 linear 953g 953g 99% | `- loop10 linear 49m 50m 98% | `- ubuntu--vg-root linear 50m 915g 0% | `- nvme0n1p2 linear 915g 953g 96% | `- nvme0n1 linear 953g 953g 99% |- md127 linear - 0 - |- md126 linear - 0 - |- md125 linear - 0 - |- ubuntu--vg-swap_1 linear - 15g - | `- nvme0n1p2 linear 15g 953g 1% | `- nvme0n1 linear 953g 953g 99% |- nvme0n1p1 linear - 512m - | `- nvme0n1 linear 512m 953g 0% |- loop3 linear - 0 - |- loop2 linear - 0 - |- loop1 linear - 0 - |- loop7 linear - 0 - |- loop6 linear - 0 - |- loop5 linear - 0 - |- loop4 linear - 0 - |- loop13 linear - 10m - | `- ubuntu--vg-root linear 10m 915g 0% | `- nvme0n1p2 linear 915g 953g 96% | `- nvme0n1 linear 953g 953g 99% |- loop12 linear - 50m - | `- ubuntu--vg-root linear 50m 915g 0% | `- nvme0n1p2 linear 915g 953g 96% | `- nvme0n1 linear 953g 953g 99% `- loop14 linear - 143m - `- ubuntu--vg-root linear 143m 915g 0% `- nvme0n1p2 linear 915g 953g 96% `- nvme0n1 linear 953g 953g 99%
The --verbose flag adds the known device paths information:
$ om node print devs --reverse --verbose aubergine Type Parent Use Size Ratio |- drbd1 - 0 - | `- loop0 linear 0 3g - /dev/disk/by-label/pridns.fs.1 | | /dev/disk/by-uuid/27489964-94c | | 6-4803-a389-c91463da8d3f | | /dev/loop0 | `- ubuntu--vg-root linear 3g 915g 0% /dev/disk/by-id/dm-name-ubuntu | | --vg-root | | /dev/disk/by-id/dm-uuid-LVM-vz | | I1exojgdAZhf3X1Vz8A0C1Ne2EN2sr | | zZlazB8vy5ey8yftklunzMMMUxJwwC | | ej | | /dev/disk/by- | | uuid/3653539e-3299-448e-b80d- | | 576fb6b71b84 | | /dev/mapper/ubuntu--vg-root | | /dev/ubuntu-vg/root | | /dev/dm-0 | `- nvme0n1p2 linear 915g 953g 96% /dev/disk/by-id/lvm-pv-uuid- | | VilFt5-Ne8T-eVxf-QKnX-n0Zc- | | LIK2-7ct9Kr | | /dev/disk/by-id/nvme- | | SAMSUNG_MZSLW1T0HMLH- | | 000L1_S308NX0J403249-part2 | | /dev/disk/by-id/nvme-eui | | .002538b471bb6f3c-part2 | | /dev/disk/by-partuuid/c6c21095 | | -4a8e-4461-af40-12e78fd758d6 | | /dev/disk/by-path/pci-0000:05: | | 00.0-nvme-1-part2 | | /dev/nvme0n1p2 | `- nvme0n1 linear 953g 953g 99% /dev/disk/by-id/nvme- ...
Each resource driver can implement a method to claim devices as "exposed" and "sub" (exposed layered over). The device tree is walked to determine the "base" devices:
$ om pridns print devs pridns |- disk#0 (disk.loop) | |- base | | `- /dev/nvme0n1 | `- exposed | `- /dev/loop0 |- fs#1 (fs) | |- base | | `- /dev/nvme0n1 | `- sub | `- /dev/loop0 `- fs#2 (fs) |- base | `- /dev/nvme0n1 `- sub `- /dev/loop0 $ om pridns print exposed devs pridns `- disk#0 (disk.loop) `- exposed `- /dev/loop0 $ om pridns print sub devs pridns |- fs#1 (fs) | `- sub | `- /dev/loop0 `- fs#2 (fs) `- sub `- /dev/loop0 $ om pridns print base devs pridns |- disk#0 (disk.loop) | `- base | `- /dev/nvme0n1 |- fs#1 (fs) | `- base | `- /dev/nvme0n1 `- fs#2 (fs) `- base `- /dev/nvme0n1