Workload Partition (WPAR) & Workload Partition Manager - Frequently Asked Questions (FAQ)
These questions were asked on the Mid USA AIX User Conference call when WPAR was covered in three sessions. I thought they would make a good start.
Question 1. Is it possible to allocate SAN storage to WPAR?
It is not possible to have devices (disks or Ethernet) in a WPAR, in this release. You can include the LUN in the global AIX, make a file system in it and then make that file system available to the WPAR or include a SAN based file system in a WPAR via NFS.
2009-10-20 Announced support for SAN, see http://www-01.ibm.com/common/ssi/cgi-bin/ssialias?subtype=ca&infotype=an&appname=iSource&supplier=877&letternum=ENUSZP09-0370&open&cm_mmc=5763-_-n-_-vrm_newsletter-_-10285_134933&cmibm_em=dm:0:10055635
[EMEA version].
Question 2. Errpt of WPARs ... where will they show up?
Each WPAR has its own errpt details - these come from its own private /var files. In a freshly started new WPAR I find for example:
# errpt
IDENTIFIER TIMESTAMP T C RESOURCE_NAME DESCRIPTION
A6DF45AA 0801090108 I O RMCdaemon The daemon is started.
9DBCFDEE 0801090108 N U errdemon
Question 3. Third party certifications? Is there a URL?
IBM does not certify or track third party product certifications - there are tens of thousands of them. It is expect most applications will just run in a WPAR without any changes. You should ask your ISV about WPAR support. This number is growing as AIX 6 (including the open beta program for ISV testing) has been out since June 2007. Asking them might prompt them to go take a look at both AIX 6 and WPARs (if they haven't already) which should be encouraged earlier rather than later.
Question 4. If /usr for regular WPARs is just a link to the global AIX copy, why are devices and bos filesets being applied?
Each WPAR will have its own /etc and /var and other files that make up parts of AIX. These are built by the installation of the right filesets. The shared /usr is (or should be) static and read only and large in comparison. AIX is something like 2 GB but by sharing /usr and /opt each WPAR size comes down to 50 to 60MB in size.
Question 5. Is the WPAR a System WPAR by default?
No.
- If you use the mkwpar command you make a System WPAR.
- If you use the wparexec command you make an Application WPAR. Note that this command includes the mandatory program/command that makes the core of the Application WPAR.
Question 6. If I am creating an application WPAR for say running the DB2 database instance, does the application WPAR die if the command to start DB2 fails?
Fails! Fails! Why would it fail? An application WPAR will stop once the program you start for the WPAR to run completes and all sub processes that it started finish too. But note: DB2 is an ideal application to run in a System WPAR - you can make the data only available to the WPAR and you have extra security plus you can monitor just the WPAR/DB2 work.
Question 7. Do all the WPARs have to be on the same subnet?
Yes and no. If your global AIX has access to multiple subnets then the WPARs that it hosts could be assigned to any or all of them. If you want to Live Application Mobility a WPAR to a different global AIX then the target also needs the same subnet.
Question 8. How do I tell how many WPARs exists from global AIX environemnt?
Use the lswpar command and you get for example a list like this:
# lswpar
Name State Type Hostname Directory
-----------------------------------------------
my_wpar A S my_wpar /wpars/my_wpar
wp01 D S wp01 /wpars/wp01
wp05 A S wp05 /wpars/wp05
wp06 A S wp06 /wpars/wp06
wp07 A S wp07 /wpars/wp07
wp08 A S wp08 /wpars/wp08
wp10 A S wp10 /wpars/wp10
wp13 A S wp13 /wpars/wp13
Key:
- A = Active = running
- D = Defined = not running but could
- S = System WPAR
Question 9. Is a WPAR's root user the same as the global root user?
No. The system WPAR has its own unique root user, password and other users - this extra security level is the point of having WPAR i.e. root users that are separate. The WPAR root user can only affect the WPAR's files and processes - i.e. they can't effect change in the global AIX level - this allows the global AIX root user to hand out some tasks without risking the global AIX security.
Question 10. How big are WPARs because they might cause problems & fill up rootvg - depends on how much stuff you put in the WPAR file system?
This is something to watch out for but that is normal AIX system administration. Note: in the current release, a WPAR can't access a volume group - only the file systems. So I guess your question applies to file systems filling up. By default, each WPAR gets its own root part file systems anyway, so one WPAR filling its file systems won't affect any other WPARs. The maximum amount of disk space a WPAR can consume is therefore limited to the size of those file systems. If you are concerned about filling up rootvg, you can specify an alternative volume group using "-g othervg" to mkwpar.
Question 11. Are there still soft & hard limits?
Yes, there are resource limits for CPU and memory use. It uses the same mechanism as for Workload Manager - a tool available in AIX for ten years or more. There is also a theoretical limit on the number of WPARs in global AIX of a little over 8000. (You are likely to encounter practical limits much sooner due, for example, to disk space.)
Question 12. Can WPARs use Veritas Cluster Server (VCS) or use HACMP cluster?
Not directly in the WPAR but the global AIX level could use VCS for file systems that are allocated to the WPARs and HACMP for DR. You really need to ask the vendor as only they decide if it's supported or not for WPAR use.
Question 13. Is the memory assigned through Workload Manager (WLM)?
By default all WPAR processes compete for memory resources just like processes do in AIX - because that is what they are. WLM like control can also be used to monitor and control memory use. Be careful, controlling memory is an easy way to cause performance problems rather than solve them - you can generate completely unnecessary paging which really hurts performance, so try this gently after monitoring it.
Question 14. Can you increase/shrink a file system on the fly?
The underlying JFS2 file systems at the global AIX level can be changed on the fly - no difference for WPARs use.
Question 15. Would the WPARs be backed up in a mksysb of the global environment?
Yes and no!
- To be included in a mksysb the file systems have to be in rootvg AND to be online/mounted (WPAR running).
- I would recommend placing WPARs outside the rootvg volume group or on NFS for Mobility/Relocation. Also note if the WPAR is not Active i.e. running then the file system is unmounted and would not get on to a mksysb at the global AIX level.
- I would recommend backing up at the WPAR level. Why? For a shared /usr and /opt the WPAR is very small and quick to backup with the WPAR commands - this makes it independent of the global copy of AIX i.e. you could restore it on a different global AIX.
Question 16. Can a WPAR be created on the SAN disk?
WPARs don't have devices like disks. You can have the SAN disk at the global AIX level. Create a file system in it and mount the file system into the WPAR. Or have the SAN on the NFS server and NFS mount the file system to the WPAR.
Question 17. Can you use Etherchannel and network backup for the WPAR?
WPARs don't have devices like networks. It would make sense for the global AIX to use Etherchannel for its network adapters then all WPARs would benefit with faster networks and recovery.
Question 18. How do you install applications to a single WPAR?
- If they don't install into /usr or /opt, or touch those file systems after installation, you can install them into the WPAR as normal.
- If they install into /usr and/or /opt, but do not touch those file systems after installation, you must first install them into the global and then execute syncwpar on all "shared" WPARs (those with read-only /usr and /opt) in which you wish to execute the application.
- If they do touch files in these two file systems after installation (tut tut) then create your WPAR with private /usr and /opt (see mkwpar -l) and install them inside the WPAR.
- An alternative for applications that only affect a self-contained subdirectory within one of these two file systems is to just add that directory as a read/write file system. The /usr/ and /opt can remain read-only. See the next question.
Question 19: My application writes to a sub-directory of /opt but I don't want a private /opt. How can I do this?
You can add a read/write directory to the read-only shared /opt - same goes for /usr. For example: creating a read/write /opt/new directory in the WPAR called "wp01":
# lswpar
Name State Type Hostname Directory
-----------------------------------------
wp01 A S wp01 /wpars/wp01
Create a mount point in the global /opt so that it appears in the WPAR /opt
Use the global AIX read-write copy as the WPAR one is read-only
# mkdir /opt/new
Create a filesystem
# /usr/sbin/crfs -v jfs2 -g rootvg -m /wpars/wp01/opt/new -a size=10M
File system created successfully.
130864 kilobytes total disk space.
New File System size is 262144
Mount it into the WPAR
# mount /wpars/wp01/opt/new
Now check it works OK by entering the WPAR and writing a file
# clogin wp01
*******************************************************************************
* Welcome to AIX Version 6.1! *
* Please see the README file in /usr/lpp/bos for information pertinent to *
* this release of the AIX Operating System. *
*******************************************************************************
# touch /opt/new/freddy
# ^D
# ls /wpars/wp01/opt/new
freddy lost+found
Question 20. Can WPAR file systems be created outside of rootvg?
Sure. It is just a file system. The demo machines used to create the WPAR Movies tend to be a simple setup and a "crash and burn" scenario so they don't have lots of volume groups. A simple and sensible way is to create a /wpars file system in any volume group you like.
Question 21. Assuming WPAR file systems can reside outside of rootvg, does a system WPAR need to be backed up using the savewpar vs. mksysb command?
Use the WPAR commands to save and restore individual WPAR and given a shared /usr you will be amazed how small they are.
Use mksysb to backup the global AIX - which should not be changing much - treat it as a bucket to run WPARs and not as a general purpose AIX which also does a bit of WPARs. Also see question 15.
Question 22. For the non-shared configuration, is GPFS supported?
WPARs don't have devices like disks. GPFS is not supported in this release. I am told you can NFS export a GPFS file system. NFS is supported so GPFS could be used to make the file systems redundant and NFS for actual access.
Question 23. If you have multiple network adapters in the global instance, can different subnets be used? If so, how can you distinguish which adapter is used within a WPAR?
Yes, the global AIX can have multiple networks and you can specify which you want the WPAR to join as part of the network parameters. Demo system tend to have just one network to keep it simple. To find out more read the manual for the mkpar command - yes it is a very complex/power command so look for example a command like: mkwpar -N interface=en4 ...
Question 24. While the application was running is it possible to log into it?
Assuming an Application WPAR, yes even this simpler WPAR type is a full WPAR. However, since application WPARs don't run system services such as inetd, it is not possible to telnet to them. The clogin command is available for access, but it is not recommended for general use. There should be no need to log into an application WPAR, though. Most information of interest can be obtained from the global. For example, to list the processes inside the wpar, you can specify "-@ wparname" to the ps command.
Question 25. Is disk cache shared among the wpars?
Yes, the AIX file system cache operates for global AIX I/O and WPAR I/O.
Question 26. Can I redirect the location of savewpar to another location, instead of /tmp/wpardata ?
No, but in this directory are a set of temporary only files about the setup and content of your WPAR. These files are then actually backed up to /dev/rmt0 by default but you can change that to other tapes or to a file in a file system with the -f flag. So /tmp/wpardata is just a useful place to find your WPAR details.
Question 27. How and where do I patch in this environment?
Patch!! - our products are perfect why patch them? 
In the unlikely event of a patch, you would patch the global AIX level - some require a reboot to activate the new AIX kernel and that will not change - and then execute syncwpar -A. If you have a private /usr or /opt file systems in one or more WPARs, you should use the syncwpar command with the -D flag to make sure they are also updated with any patches in /usr or /opt.
Question 28. Is a system WPAR root user the same as the global AIX environment root user?
No. There is a separate root privilege set for each system WPAR. So it can have a root user and unique password that only has root privileges for that particular WPAR - not the whole global AIX instance (LPAR).
Question 29. Can a global AIX environment be a regular LPAR and does the WPAR share the root file system?
The global AIX environment is the AIX running in a machine or a logical partition. Each WPAR has its own top level root file systems called "/", "/tmp", "/home", and "/var". The WPAR definition decides what sub-directory or NFS mount point is going to be used for the top level directory for the WPAR. Note: in the WPAR spec file "/" refers to the WPAR's "/" and not the global AIX "/".
Question 30. How would Partition Mobility work if the WPAR is part of the local root volume group?
With Partition Mobility (PM), the whole AIX image (what in WPAR terms many people also call the global AIX level) is moved to the new POWER6 machine. This will include all the WPAR details and content. Partition Mobility does this by moving the SAN disk connection and Network (both of which have to be via the Virtual I/O Server) to the new machine. For example, a WPAR using a file system in the rootvg or other volume group would move with the LPAR.
Question 31. Do AIX TL updates get shared to the WPARs when applied to the host system?
Yes, once syncwpar has been executed, since it is all one AIX instance if using the regular shared /usr and /opt which is normal. See Question 27.
Question 32. Can you run a 32 bit kernel in an WPAR?
No, AIX 6 is a 64 bit kernel only - who would want a 32 bit kernel anyway? The 64 bit AIX kernel runs 32 bit AND 64 bit application all at full speed.
Question 33. If the application WPAR shares the global file systems, what do you gain?
You have encapsulated your workload; now it can be monitored and controlled. If you have WPAR manager you can also pause, suspend, checkpoint and restart your application WPAR. You can also use Application Mobility to move your application WPAR to a different machine. You also have fewer AIX images to manage so you save a lot of man power.
Question 34. What would an example of an application WPAR usage? Just trying to understand what you gain by running one.
I see High Performance Computing (HPC) being a obvious candidate for a application WPAR. These are often programs (or collections of processes) that run for days or weeks and the encapsulation would be excellent (see above question). Second would be a application that has no data (so there is no real need to have separate file systems) i.e. a network task or providing some compute-only task.
Question 35. Can he explain the wparexec command parameters?
See the IBM website for the AIX manuals: There are many options. Briefly:
# wparexec
Usage: wparexec [-a] [-c [-1]] [-F] [-h hostName] [-M attr=value ...] ... [-n wparName]
[-N attr=value ...] ... [-R attr=value ...] [-u userScript] [-v]
{ [-e existing_wpar | -f inFile] [-o outFile [-w]] [--] [VAR=val ...] Command [args]
| {-f inFile | -e existing_wpar} [-o outFile [-w]] [[--] [VAR=val ...] Command [args]]
| -w -o outFile [-f inFile | -e existing_wpar] [[--] [VAR=val ...] Command [args]] }
Flags:
-1 = Create configuration file only
-a = Automatically resolve erroneous/conflicting settings.
-c = Flag as checkpointable.
-e = Get specification data from existing workload partition.
-f = Path to input specification file.
-F = Force - ignore certain errors.
-h = Hostname.
-M = File system dependencies (dev, directory, vfs, mode, host, mountopts).
-n = Workload partition name.
-N = Network settings (interface, address, netmask, broadcast).
-o = Path to output specification file.
-R = Resource control settings (active, rset, CPU, memory, procVirtMem,
shares_CPU, shares_memory, totalProcesses, totalThreads).
-u = User script to execute on start & stop.
-v = Verbose mode.
-w = Only write specification file (do not create).
Question 36. If the global environment file system is mirrored, doesn't it logically follow that any data from the WPAR placed on that mirrored file system is also mirrored because it went onto a mirrored volume?
Absolutely.
Question 37. How is DB2 and Oracle licensing governed in WPARs?
Interesting question and one for the vendors.
The below is a personal comment:
- I find the ISV's (and only with their latest releases) barely coping with technology from 2 years ago like shared CPU logical partitions! The Systems Administrator can limit a WPAR to a percentage of the machine/LPAR or a resource set like particular CPU but it may be a software vendor release cycle or two until this is fully handled. The important thing for now is to point out that you want WPAR and better license support from your vendor, so at least they start thinking about it and planning for it.
Question 38. Are the file systems within a WPAR stored as large files on the LPAR? Thus if you have a 4G file system on a WPAR, it would show up as a large 4G file within the LPAR?
No, the file systems in the WPAR are file systems in the global AIX too. This means the global AIX can "see" all the files in the WPAR and can, for example, copy a file to a WPAR's file system just using the "cp" command.
Question 39. Can a system WPAR have a different root file system than the LPAR it is on?
The WPAR MUST have a different root file system to the global AIX root file system. This is not optional. In the global AIX this is mounted at /wpars/<wparname>/ by default. This file system contains, for example, the /etc files which must be unique to the WPAR - this forces the fact that each WPAR has its own root file system for its own files. Only /usr and /opt out of the global AIX file systems directories are normally shared read-only; other less critical file systems such as /home and /tmp can be shared too for read-only or read-write access.
Question 40. When you stop a WPAR, will it run stop scripts?
You can stop a WPAR in at least three ways
- logon and run shutdown,
- the stopwpar command from the global AIX level and
- a stop from the WPAR Manager GUI (actually uses stopwpar under the covers)
All these will run the WPAR script "/etc/rc.shutdown" within the WPAR.
Question 41. Is there a Command Line Interface (CLI) for all the settings visible in the WPAR Manager or just a subset?
The Graphical User Interface (GUI) uses the CLI under the covers. All of these are available to users. The "chwpar" command, for example, is used to change many WPAR settings.
Question 42. *Can I use NIM to deploy the WPAR Manager agents?".
The WPAR Agent and MCR (checkpoint/restart) AIX LPPs packages are regular AIX style installp packages and so can be remotely installed by NIM to the global AIX. You don't install the agents to the WPARs themselves. You can also include the packages in your standard mksysb image for new installs of a global AIX.
Question 43. Is WPAR Manager required for relocating a WPAR?
Workload Partition Manager gives you the Manager GUI and internal database, the Manager Agents and the two commands /opt/mcr/bin/chkptwpar and /opt/mcr/bin/restartwpar. Without the these two commands WPAR Relocation (also called Application Mobility - it's the same thing) is impossible, although you could backup, move the file, remake the WPAR and restore by hand for a cold move.
Question 44. I am curious why the checkpoint/restart commands are only available when WPAR Manager is purchased?
If you want the function legally, you need to purchase the license. That is just the way it is. The costs are relatively small but as "techies" we don't discuss prices on Wikis or Forums. Having worked with tens of WPARs on my demo systems, I would have gone mad trying to keep track of all the WPARs without WPAR Manager and would have had to play "hunt the WPAR" to find WPARs that I have moved around but forgotten where I put them last. The advanced Policy based Relocation functions are also important. I tend to create WPARs from a script but everything else I do is via the WPAR Manager. Oh except backup and restore. I recommend you don't try WPARs without the manager.
Question 45. When you move a WPAR, can you move the entire global environment too?
When a WPAR is moved, it is only the WPAR processes' memory and state and files that are moved to the target machine. The global AIX stays put and the WPAR files are moved via reconnecting NFS. This means only the WPAR processes' memory needs to capture and the associated kernel details like open files, sockets and settings - which makes it quite small on a simple WPAR. This is not to be confused with Live Partition Mobility (PM), a very different technology, that moves the entire LPAR contents to a new POWER6 machine.
Question 46. Please explain the relationship to Live Partition Mobility functions?
- Live Partition Mobility (see above question) requires very different pre-reqs: POWER6 hardware with AIX 5.3 (recent TL), AIX 6 or Linux (SLES or RHEL) and the LPAR must be using only virtual disks and virtual Ethernets via the Virtual I/O Server.
- Live Application Mobility only requires AIX 6, which in turns needs a POWER4, POWER5 or POWER6 machine.
The two mobilities are totally different technologies.
Question 47. If you are familiar with Solaris containers/zones, can you compare similaries/differences to WPAR technology?
IBM, of course, does not like commenting on competitor solutions as they are not experts on it. Sure they are aware of the technology and its features and think WPAR is far better. WPAR is not new technology as IBM has been running many years of beta tests with selected customers before the initial release. Certainly, zones can't do anything like Live Application Mobility and it is a complex technology to get right so Sun is unlikely to suddenly add this feature. IBM bought a French company to gain the technology, patents and expertise to add this to AIX 6.
Question 48. Where can you learn more information about WPAR states?
Question 49. Will WPARs be supported in AIX 5.3 at some point (since it's completely SW based)?
No, WPAR support means large changes to the AIX kernel, including additions like RBAC. Given this it does not make sense to add this to AIX 5.3 as it would require ISV recertification. AIX 6 really does make sense.
Question 50. Which browsers are supported with WPAR Manager?
Internet Explorer and Firefox. Supported browser versions can be found at http://publib.boulder.ibm.com/infocenter/systems/scope/aix/topic/com.ibm.aix.wparlpp/wparlpp.pdf
Currently, only these two browsers are supported - if you really can't use these get in touch. I think this is a testing issue i.e. IBM is not going to test the other dozens of browsers. Don't forget you can run Firefox on AIX! I have been using Firefox 3 for some time, not officially supported but works fine.
Question 51. Is WPAR Manager agent included in AIX 6 by default?
Nope, it comes with the WPAR Manager, which needs to be purchased/licensed.
Question 52. How much server resource - CPU/Mem is required to run the agent and what is the best capacity planning tool that is fully supported?
I found the prerequisites at http://publib.boulder.ibm.com/infocenter/systems/scope/aix/topic/com.ibm.aix.wparlpp/wparlpp.pdf
The official answer is WPAR agent = 45 MB memory when idle + 27 MB disk space per global AIX (not per WPAR).
I can only barely detect the agent using CPU on my global copies of AIX - I guess it will use CPU while Relocating a WPAR or at least the WPAR commands it used underneath will.
Question 53. Can Workload Partition Manager be installed on/in a WPAR?
Absolutely not - nice try! WPARs have no disks or adapters and only file systems to get to data, so they can't be nested.
Question 54. When relocating a WPAR, does it remove the filesets from the source host?
Filesets are not effected on the global AIX, the WPAR Agent and MCR is installed once and stays with the global AIX.
Question 55. So would it be safe to say that all global environments must be on the same subnet, if the WPARs are to be mobile since network aliases are used?
Yes, you can shutdown a WPAR cold move it, fix the WPAR network details, and start it. Not Live Application Mobility but still only a couple of minutes work. Don't forget the global AIX can have multiple networks; join the WPAR to the right ones and hopefully you can do Live moves by having all global AIX images on all required subnets.
Question 56. * Can you move between POWER5 and POWER6?*
You can't do Live Application Mobility between CPU types - the problem is the target CPU may not have the same instruction set and so programs that learn the actually CPU and optimize for it would (possibly) crash. But you can still move the WPAR (assuming NFS file systems) with WPAR Manager:
- Stop the WPAR on the source global AIX
- Delete the WPAR but keep the WPAR definition (this is called a Undeployed state)
- Deploy the WPAR on the target global AIX
- Start the WPAR
This means a few minutes (like 1 or 2) downtime. Future CPUs may have emulation of older chips built in so then it could go from POWER N to POWER N+1 (but running in POWER N mode). This is not an announcement!
Question 57. Is there a setting to make a system WPAR auto restart with the global AIX start?
Yes there is an option for each WPAR to auto start when its current global AIX starts up (after a failure or user activated reboot). On the command line, this is the -A option to the mkwpar command.
Question 58. Is there a dedicated network requirement to complete a relocation? Like a dedicated or private Network VLAN.
The source and target global AIXes have to be on the same subnet. The process memory and state data is moved between them by putting it on the NFS file system that they both have access to. There is no global AIX to global AIX data move. A faster method may be available in the future.
Question 59. Is there a network outage during a relocation?
The WPAR processes are frozen so it depends on how applications are handling this. Most user applications will see it as a temporary network glitch and are built to timeout and retry. Applications such as browsers can be handled by users retrying but if your application can't work through a few packets going missing (this can happen on any network) then you may have problems. I note that a common and really useful program called "putty" instantly detects the other WPAR end is not present and closes the connection - to be honest this is a pain in the neck. Where as telnet from AIX just carries on with no issues.
Question 60. High availability: If a host fails and hence the WPAR goes down, can you manually start that WPAR on another available LPAR? .
Yes, the WPAR Manager will notice the WPAR Agent go missing in action and highlight this on the GUI. The WPAR state goes to "unknown". I have not tried it but I assume you can run a forced "stopwpar", cold relocate the NFS file systems to an alternative global AIX and restart it.
Question 61. What are the WPAR commands to manualy Relocate?
Here is a worked example - you will have to use your WPAR name (wp03) and network/NFS details:
On the source this checkpoints and kills (-k) the processes afterwards:
/opt/mcr/bin/chkptwpar -d /wpars/wp03/tmp/state -o /wpars/wp03/tmp/checkpoint.log -k wp03
On the target define the WPAR - this is the same as the WPAR create command but the -p means preserve the NFS files content and don't install:
mkwpar \
-n wp03 \
-h wp03.aixncc.uk.ibm.com \
-N netmask=255.255.255.0 address=9.69.44.113 \
-r \
-c \
-M directory=/ vfs=nfs host=9.69.44.249 dev=/nfs/wp03root \
-M directory=/home vfs=nfs host=9.69.44.249 dev=/nfs/wp03home \
-M directory=/tmp vfs=nfs host=9.69.44.249 dev=/nfs/wp03tmp \
-M directory=/var vfs=nfs host=9.69.44.249 dev=/nfs/wp03var \
-p
Then restart from the saved files to recreate the processes i.e. WPAR:
/opt/mcr/bin/restartwpar -d /wpars/wp03/tmp/state -o /wpars/wp03/tmp/checkpoint.log wp03
On the source, clean up - remove the WPAR definition:
Question 62. Can you run the partition manager software and the agent on the same machine in the same LPAR? or does the manager software need to run on its own machine?
Yes, you can run WPARs on the same machine or LPAR as the WPAR Manager - it is just regular AIX running a little extra software as an application. The installed WPAR Manager software includes the WPAR Agent.
Question 63. What is the purpose of the WPAR group column?
A WPAR can placed in to WPAR Groups. This can help in organizing the WPAR Mgr screens. You can also add thresholds for a group and a Relocation policy and then the WPAR Manger will automatically move WPARs around within the group to optimize performance. You do have to "let go" and trust the WPAR Manager and it does not instantly start move WPARs - That would not be a good idea - we don't want to take even small outages due to a tiny, say one second, peak. But over say, 10 to 15 minutes, you find the WPAR Manager will start making changes. Note: though that its not a good demo as it feels like it takes a very long time - if you sit and watch it.
Question 64. What is the state of the AIX processes during the relocation?
No change but WPAR processes are frozen.
Question 65. What should not be done at the global environment level when WPARs are there?
The WPAR is just a bunch of processes with an extra flag as far as the global AIX in concerned so its just the same as for a regular application - they need disk and network access and some CPU and memory. But I have had problems Relocating if on one machine I have tried security features like Role-Based Access Control (RBAC). This adds special kernel extensions and changes to the global AIX environment. So during Relocation it fails to build a duplicate environment ( one has the extensions and one doesn't) and the Relocate has to be backed out so the WPAR restarts again on the source LPAR.
Question 66. Can you have a single shared file system across two or more WPARs that are located on the same machine/LPAR ( not using NFS )?
Yes. For example, if you have a filesystem called /local and what to add this to two WPARs called wp20 and wp21 then assuming you have created the file system and have it mounted, on the global AIX level as root:
mkdir /wpars/wp20/local
mkdir /wpars/wp21/local
mount -v namefs /local /wpars/wp20/local
mount -v namefs /local /wpars/wp21/local
You will now find the global filesystem /local is also present in the two WPARs.
Question 67. Should I use the global AIX level for non-WPAR work?
If you want to avoid WPAR Relocation failures - Absolutely Not. Some minor "other work" is OK but anything that may involve an AIX kernel extension or additional libraries should be avoided. It is difficult to know if using or installing some features will have these unpleasant side effects so the best practice is
- Keep the global AIX images identical and simple
- Don't do anything at the global AIX level - even logging in as root should be avoided whenever possible to limit accidental modifications. OK you may have to for WPAR administration but try to minimise root use.
- Don't add any extra software, install new packages or "experiment" at the global AIX level unless you a) know what you are doing and b) intend to duplicate the same change to all global AIX images.
To give you an example: I tried the Encrypted File System features of JFS2 on one global AIX. This involved switching on the EFS feature and unknown to me this adds a kernel extension to handle access to the encrypted files. A few days later my next WPAR Relocate fails in fact any Relocate from this global AIX failed. I did not connect this to the EFS work at first. Digging through the Relocate error logs, I found references to missing kernel extensions when trying to restart the WPAR on the target (unmodified) global AIX. To get out of the issue I could use one of the three methods below:
- Activate EFS on all my global AIX images (it was only an experiment so no thanks)
- Uninstall EFS from my global AIX. I tried that but the global AIX is then still slightly different - it knows about the EFS kernel extension and still causes Relocate failures as the source and target global AIX's are still slight different.
- Reinstall AIX into the global AIX. Fortunately my WPARs where all NFS mounted, so I could stop them, undeploy them and then reinstall a fresh AIX in the global AIX LPAR. Then I redeployed the WPARs back and start them again on the global AIX. With WPAR Manager and reinstalling via NIM this only took 15 minutes. But I learnt my lesson = don't fiddle with global AIX images.
Assuming WPAR Relocation is important to you - Think of the global AIX as an "empty shell" or "bucket" used to host WPAR and not to be used directly for running workloads. Don't treat them a general purpose AIX.
Question 68. Did you know about the WPAR Manager automatic hostname lookup and list?
When you are typing a WPAR system hostname into a WPAR Manager field that is will automatically list hostnames that match the current input: For example, (see below) I typed "gold" and paused for a second. The WPAR Manager pops up a list of the hostnames starting with "gold" for me to select from. This saves a lot of typing and removes many typing mistakes.
Question 69. How can I get hold of a WPAR specification file?
I can think of number ways to do this
- When you create a system WPAR with the mkwpar command you can add the "-o filename" to save a copy for future use
- Take a look in /etc/wpars and you should find a "wparname.cf" file for each of the WPARs on the system - as far as I know you should not meddle with this file.
- If you run the command "mkwpardata wparname" then it will create a directory "/tmp/wpardata/wparname" in which you will find a file called wpar.spec. The mkwpardata command creates the data files used to backup your WPAR with savewpar and to restore the WPAR from scratch with the restwpar command.
- You can clone a current WPAR without using a spec file. To do this you can use mkwpar with the "-e wparname" option. This will use the named WPAR to set all the options, which you can then overwrite with other options like the WPAR name, network details etc.
Question 70. Can you give an example, of how to backup, restore and clone a WPAR ?
Assuming you are using a NFS based WPAR so that you can use Application Mobility (Relocation), in the global AIX environment as root user with the WPAR called wpar123 running, use the command:
savewpar -Nif /backup/mybackup.bk wpar123
You can call the backup file any thing you want and it is in backup/restore format. This on a simple default WPAR just install only takes 20 seconds and results in a file ~70MB file. The -N option, includes the NFS mounted files - this is very important!
To recover the WPAR from a disaster, first completely remove the WPAR (rmwpar) and delete the files in the NFS mount points as it will get recreated from scratch then
restwpar -f /backup/mybackup.bk
If you have only lost a few files, you can also recover individual files from the backup without this destroy and recreate process.
Cloning can be viewed two ways
- A restore but change a load of WPAR details beforehand rather than using the contents of the backup for the WPAR details
- A make WPAR command but initialise the WPAR from the backup file rather than installing it
Here is an example:
restwpar -h wpar321 -n wpar321 -d /wpars/321 -r -U \
-M "-N address=99.55.44.33 \
-M directory=/ vfs=nfs host=gold dev=/nfs/wpar321root \
-M directory=/home vfs=nfs host=gold dev=/nfs/wpar321home \
-M directory=/tmp vfs=nfs host=gold dev=/nfs/wpar321tmp \
-M directory=/var vfs=nfs host=gold dev=/nfs/wpar321var" \
-f /home/mybackup.bk
This sets a different WPAR name, hostname, top directory, network address and NFS mount points.
Note the double quotes around the -M option. This can be quite tricky to get right!
