There are two ways to manage power in RHV. The first is the traditional way that we’ve all used for years, which is a set of power profiles. However, RedHat’s new Virtualization Manager (VMM) can make managing these power profiles much easier. For example, if you’re an Enterprise customer (or plan to be), you can easily create a set of power profiles and assign them to machines based on tags. This makes managing your servers much easier than ever before.
I was working on an article recently for one of my employers about the topic of power management in RedHat Virtualization (RHV) and I was having a lot of problems with getting it done. I was trying to find something specific about power management in RHV and this is what I have found, not only did I not find anything about the topic I was looking for, but I also found a lot of information on various topics. Unfortunately I was unable to get the information I wanted so I decided to share what I did find.
If you are a RedHat user, you may have come across the concept of power management in virtualization. You may have also heard of kvm, vxlan, and vlan interfaces. These are all names for the same thing – Power management in virtualization. I will explain all of these concepts, and explain what they mean, and why they are important, in this blog.. Read more about rhv power management and let us know what you think.
This article will delve into RHV Power Management in depth.
What is Host Power Management (Fence) and how does it work?
RHV can reboot hosts that are in a NonOperational or NonResponsive condition when Power Management is enabled. The following power management devices are supported by RHV:
- Conversion of American Power (apc)
- Bladecenter by IBM (Bladecenter)
- Cisco Unified Computing System (cisco ucs) is a computer networking system developed by Cisco.
- Remote Access Card for Dell 5 (drac5)
- Remote Access Card for Dell 7 (drac7)
- Power Switch with Electronics (eps)
- BladeSystem from HP (hpblade)
- Integrated Lights Out (ilo, ilo2, ilo3, ilo4, ilo ssh) (ilo, ilo2, ilo3, ilo4, ilo ssh) (ilo, ilo2, il
- Interface for Platform Management Intelligence (ipmilan)
- Adapter for Remote Supervisor (rsa)
- RSB Fujitsu-Siemens (rsb)
- Western Telematic, Inc. is a company that specializes in telematic (wti)
RHV communicates with power management devices via fence agents.
What is Auto Fencing and How Does It Work?
When a host fails unexpectedly, the host status changes to connecting, and the host remains in this state for a grace time. If the timeout period expires, the Host will become ‘NonResponsive’ or ‘NonOperational.’ Engine fences the troublesome hosts by rebooting in response to this status. Engine stops the host, confirms that it has stopped, starts the host, and confirms that it has been started using the fencing agent for the power management card on the host.
Grace Period for Auto Fences:
Engine will ask vdsm for status twice by default:
default value | option name | option value ————————————-+———————+———————- VDSAttemptsToResetCount | 2 | 2 VDSAttemptsToResetCount | 2 | 2 VDSAttemptsToReset (1 row)
DelayResetPerVmInSeconds*(Number of VMs on host) + DelayResetForSpmInSeconds = TimeoutToResetVdsInSeconds + DelayResetForSpmInSeconds (isSPM)
For instance, if the Host is SPM and has two VMs with default value, the grace period is 60+ 0.5*2+20=81s.
default value | option name | option value ——————————————+———————+———————- DelayResetForSpmInSeconds | 20 | 20 DelayResetPerVmInSeconds | 0.5 | 0.5 VDSAttemptsToResetCount | 2 | 2 TimeoutToResetVdsInSeconds | 60 | 60 DelayResetForSpmInSeconds | 20 | 20 DelayResetPerVmInSeconds | 0.5 | 0.5 VDSAttemptsToResetCount | 2 | (4 rows)
In the event of a crash, enabling “Kdump integration” simply delays the hard-fence until the host finishes writing its memory dump.
This can be set up at the Cluster Level:
AdminPortal–>Compute–>Cluster–>Edit Cluster–>Fencing Polciy–>Enable Fencing
Engine uses “SSH Soft Fencing” to try to restart VDSM through SSH on “non-responsive” hosts before “rebooting” the host.
version | option name | option value | default value ———————————-+————————————————————————-+————————————————————————+————- /usr/bin/vdsm-tool service-restart vdsmd | SshSoftFencingCommand | /usr/bin/vdsm-tool service-restart vdsmd | 4.3 (1 row)
Soft-fencing over SSH can be used on servers that aren’t set up for power management. This is not the same as “fencing.” Fencing is only possible on hosts that have power management enabled.
Choosing a Proxy
Cluster, dc is the default Power Management Proxy Preference. “other dc” can be added as an option. It will locate a proxy server with a “UP” state.
# engine-config -g -g -g -g -g – FenceProxyDefaultPreferences cluster,dc version: general FenceProxyDefaultPreferences: cluster,dc version: general
Flow of the Engine:
Meta-data for configuration
VdsFenceType,VdsFenceOptionTypes,VdsFenceOptionMapping,FenceAgentMapping, and FenceAgentDefaultParams have the following meta-data:
-[ RECORD 1 ] -[ RECORD 2 ] -[ RECORD 3 +————————————————————————————————————————————————————— apc,apc snmp,bladecenter,cisco ucs,drac5,drac7,eps,hpblade,ilo,ilo2,ilo3,ilo4,ilo ssh,ipmilan,rsa,rsb,wti version | 4.3 -[ RECORD 2 ] -[ RECORD 3 ] -[ RECORD 4 -+—————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————
-[ RECORD 3 ] -[ RECORD 4 ] -[ RECORD 5 -+————————————————————————————————————————————————————— option value | VdsFenceOptionMapping option name | VdsFenceOptionMapping option name | VdsFenceOptionMapping option_
bladecenter:secure=secure,port=ipport,slot=port; cisco ucs:secure=ssl,slot=port; drac5:secure=secure,slot=port; drac7:;eps:slot=port; apc snmp:port=port,encrypt options=encrypt options; apc snmp:port=port,encrypt options=encrypt ilo:secure=ssl,port=ipport; ipmilan:; ilo2:secure=ssl,port=ipport; ilo3:; ilo4:; ilo ssh:port=port; rsa:secure=secure,port=ipport; rsb:;wti:secure=secure,port=ipport,slot=port; wti:secure=secure,
value default | bladecenter:secure=secure,port=ipport,slot=port; cisco ucs:secure=ssl,slot=port; drac5:secure=secure,slot=port; drac7:; eps:slot=port; hpblade:port=port; ilo:secure=ssl,port=ipport; ilo2:secure=ssl,port=ipport; il
-[ RECORD 4 ] -[ RECORD 5 ] -[ RECORD 6 -+——————————————————————————————————————————————— drac7=ipmilan,ilo2=ilo default value | drac7=ipmilan,ilo2=ilo option name | FenceAgentMapping option value | drac7=ipmilan,ilo2=ilo
-[ RECORD 5 ] -[ RECORD 6 ] -[ RECORD 7 -+———————————————————————————————————————————————- FenceAgentDefault | option name option value | drac7:privlvl=OPERATOR,lanplus=1,delay=10;ilo3:power wait=4;ilo4:power wait=4;ilo ssh:secure=1;ilo3:power wait=4;ilo4:power wait=4;ilo ssh:secure=1 default value | drac7:privlvl=OPERATOR,lanplus=1,delay=10;ilo3:power wait=4;ilo4:power wait=4;ilo ssh:secure=1;ilo3:power wait=4;ilo4:power wait=4;ilo ssh:secure=1
The following meta-data can be set up in engine-config:
# grep ‘CustomFence|CustomVdsFence’ | engine-config -a Version: general CustomFenceAgentMapping Version: general CustomFenceAgentDefaultParams: Version: general CustomFenceAgentDefaultParamsForPPC Version: general CustomVdsFenceOptionMapping Version: general CustomVdsFenceType version: general CustomFencePowerWaitParam:
Other settings (timeouts and retries):
FenceStartStatusRetries: 18 version: general # engine-config -a |grep ‘FenceStart|FenceStop’ version: general FenceStartStatusDelayBetweenRetriesInSec: 10 FenceStopStatusRetries: 18 general version Version: general FenceStopStatusDelayBetweenRetriesInSec: 10
This article broadly covered the following related topics:
- rhv power management
- rhv 4.3 fencing
- red hat virtualization
- red hat enterprise virtualization hypervisor