Hello Cassandra,
from what I've read in the scripts themselves and tested in debug mode,
I came to the following partial conclusion:
1) postscripts
- postscripts : ran in post-install
- postbootscripts : ran at the first reboot after install
They're run via a wrapper called mypostscript, which basically consists
of 3 parts:
1. var, function definitions, lib sourcing
2. a section running postscripts (# postscripts-start-here / #
postscripts-end-here delimited)
3. a section running postbootscripts (# postbootscripts-start-here
/ # postbootscripts-end-here delimited)
2) /xcatpost/ dir (on CN)
this directory holds:
- a copy of MN's /install/postscripts dir (downloaded)
- pre-created mypostscript.<node> if it exists (later renamed into
mypostscript)
- the mypostcript file, which is locally reassembled by:
- include (here doc/echo + redirect) of part 1.
- download (via curl or awk script) + patch of parts 2 and 3
in the stateful node install case, this is done by an xCAT
kickstart postinstall script (post.xcat.ng which ultimately runs it)
- this later postinstall script also creates:
- xcatpostinit1.service unit file
- xcatpostinit1 ExecStart command
- xcatinstallpost (used to run postbootscript only ?)
- xcatdsklspost : general purpose script used in many different
contexts (like an updatenode -P for instance)
- /opt/xcat/xcatinfo file which is either read or written by above
scripts
My understanding is that the /opt/xcat/xcatinfo file must at least have
XCATSERVER, INSTALLDIR, HTTPPORT for xcatpostini1.service to work
It also seems to me that after first boot, some script (which one ?)
writes REBOOT=yes into it in order not to run postbootscripts again
after a new reboot ?
Am I correct ?
What's still unclear to me (the logic is hard to follow and not
everything is shown even in debug mode) is:
in what case (conceptually, not in the code) is xcatinstallpost vs
xcatdsklspost called ?
xcatdsklspost seems to be only for diskless nodes but I don't see why it
would not be called on a statefull node as well (for an updatenode -P
for instance).
I mean all three scripts :
- /opt/xcat/share/xcat/install/scripts/post.xcat.ng called from kickstart
- xcatdsklpost
- xcatinstallpost
seems to do similar tasks, basically consisting in (re)constructing a
mypostscript file and running it
It is not clear what writes what and when into /opt/xcat/xcatinfo
Finally, my understanding is that /xcatpost/mypostscript.post is a
copy/backup (done by post.xcat.ng) of the postscripts (#
postscripts-start-here section):
- what is it used for ?
- why does an updatenode -P seems to remove it ?
Last thing : post.xcat.ng seems to call postbootscripts "prebootscripts"
and postscript "postbootscripts" (which adds to the confusion) : am I
correct ?
Thanks for your help
--
Thomas HUMMEL
On 03/11/2020 11:07, Thomas HUMMEL wrote:
On 03/11/2020 03:09, Casandra H Qiu wrote:
Have you check this link:
Hello, thanks for your answer. Yes I did read it.
after "rinstall" command , you can run command "xcatprobe osdeploy -n
<nodename>" to probe operating system provision process.
The output will show from begin of provision process until end of node
installation completion. set attribute "xcatdebugmode=1" on the site
table will show more of debug information.
I'll try this
when you deploy the VM template, How did u change the network setting?
The template has an NetworkManager auto (dhcp) profile but the node xCAT
object has been created beforehand with the correct MAC address (of the
VM instanciated from the template). Then a role in the ansible playbook
which is run to further configure the VM runs updatenode -P <node>
"confignetwork -s" to "statify" the (already correct) ip address.
I
think /opt/xcat/xcatinfo just capture some information for the CN.
Don't think will use them during the provision.
I'll check all of this. Can you tell me what /xcatpost/mypostscript.post
is ? Is just for the REBOOT/postbootscripts case ?
Basically, what I want to know is this :
all those generated or patched scripts on the CN hold some node's info
(node's name, ...) : what is the source of this info (local or from the
node object definition on the MN ?) and in what file can it be safely
removed (in order not to have the template info which will be incorrect
on the VM) ?
Thanks for your help
--
Thomas HUMMEL
_______________________________________________
xCAT-user mailing list
xCAT-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xcat-user
_______________________________________________
xCAT-user mailing list
xCAT-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xcat-user
