IGNITE on HPVM host;

[Martin Ilavsky]

Hello,

I have a virtual machine (VM) running on HPVM host. Ignite2 is an ignite server used on intranet network. Everything is working ok, I can boot-off VM from ignite2 and do a restore as needed. To draw a sketch of topology: 

    LAB
    ==={iscsi network}=====================
    |                                     |
[HPVM HOST]               [ignite2]   [client1]
    |                         |           |
    ====={intranet}========================

HPVM host        -  rx1620

ignite2,client1  -  rp7400

I need to move ignite to HVPM host.  I've installed ignite on it and tried to do a golden image installation. I'm always stuck with the following error:

Retrieving File Size.
Retrieving File (TFTP)./
PXE-E18: Timeout.  Server did not respond.
Exit status code: Invalid Parameter

But I see tftp was working in syslog:

Jun 27 17:05:53 hpvmhost tftpd[9180]: receive: RRQ /opt/ignite/boot/nbp.efi octet blksize 512 tsize 0
Jun 27 17:05:53 hpvmhost tftpd[9180]: send: OACK  blksize 512 tsize 24576
Jun 27 17:05:53 hpvmhost tftpd[9180]: tftp: client does not accept options
Jun 27 17:05:53 hpvmhost tftpd[9181]: receive: RRQ /opt/ignite/boot/nbp.efi octet blksize 8192
Jun 27 17:05:53 hpvmhost tftpd[9181]: send: OACK  blksize 8192

I did verify tftpd functionality from client1 - I was able to retrieve files without problem.  I've tried to boot with dbprofile (bootps disabled in inetd.conf) and without dbprofile (bootps enabled in inetd.conf). Still the same error.  All communication is going via intranet network, not iSCSI.  I did create second dbprofile for ignite2 and I was able to boot off it without problem.

Then I tried to boot off client1 from HPVM host and it was successful.Judging from the tests I've made it's impossible to boot off VM from the ignite if the HPVM host is the same server. Some additional information is attached in the text file.

Question is:  is it possible to boot off VM client from the ignite that is installed on the same HPVM host?

[Martin Ilavsky]

It's seems to be impossible to attach text file for some reason. I'm pasting the whole output of the file here then:

### rx1620 - HPVM host firmware

SYSREV

Current firmware revisions

 MP FW     : E.03.30
 BMC FW    : 04.01
 EFI FW    : 05.48
 System FW : 04.27

### HPVM configuration

# uname -a
HP-UX hpvmhost B.11.31 U ia64 3067848483 unlimited-user license

# hpvmstatus -Pspejbl
[Virtual Machine Details]
Virtual Machine Name VM #  OS Type State
==================== ===== ======= ========
spejbl                  10 HPUX    On (EFI)

[Network Interface Details]
Interface Adaptor    Name/Num   PortNum Bus Dev Ftn Mac Address
========= ========== ========== ======= === === === =================
vswitch   avio_lan   HPnet      2         0   1   0 7e-b4-67-e8-17-7d


#hpvmnet -SHPnet
Name     Number State   Mode      NamePPA  MAC Address    IPv4 Address
======== ====== ======= ========= ======== ============== ===============
HPnet         2 Up      Shared    lan0     0x0011855f07d0 $HPVM_HOST_IP


[Port Configuration Details]
Port    Port         Port     Untagged Number of    Active VM    Tagged
Number  State        Adaptor  VLANID   Reserved VMs              VLANIDs
======= ============ ======== ======== ============ ============ =============
1       Reserved     avio_lan none     1                         none
2       Active       avio_lan none     1            spejbl       none

### inetd.conf when dbprofile is used

#grep -vE '^$|^#' /etc/inetd.conf
ftp          stream tcp6 nowait root /usr/lbin/ftpd     ftpd -l -a
tftp  dgram  udp  wait   root /usr/lbin/tftpd tftpd /opt/ignite /var/opt/ignite
auth         stream tcp6 wait   bin  /usr/lbin/identd   identd
hacl-cfg    dgram   udp6    wait    root  /usr/lbin/cmclconfd cmclconfd -p
hacl-cfg    stream  tcp6    nowait  root  /usr/lbin/cmclconfd cmclconfd -c
hacl-probe  stream  tcp    nowait  root  /opt/cmom/lbin/cmomd /opt/cmom/lbin/cmomd -f /var/opt/cmom/cmomd.log -r /var/opt/cmom
registrar stream tcp6 nowait root /etc/opt/resmon/lbin/registrar /etc/opt/resmon/lbin/registrar
instl_boots dgram udp wait root /opt/ignite/lbin/instl_bootd instl_bootd

### inetd.conf when dbprofile is NOT used, following line is included (tested with bootpquery)
bootps dgram udp wait root /usr/lbin/bootpd   bootpd -P

### and using /etc/bootptab

ignite-defaults:\
ht=ethernet:\
hn:\
bf=/opt/ignite/boot/nbp.efi:\
bs=61

IADEFV:\
tc=ignite-defaults:\
bp=$HPVM_HOST_IP:\
sm=255.255.252.0:\
ns=$NS_IP:\
gw=$GW_IP:\
vm=rfc1048

spejbl:tc=IADEFV:ip=$SPEJBL_IP:ha=7EB467E8177D


### SW installed

#swlist |grep -E 'IGNITE|^[ ]*VM|DHCP'

 IGNITE                                C.7.12.519     HP-UX Installation Utilities (Ignite-UX)
  VMGuestLib                            B.04.30        Integrity VM Guest Support Libraries
  VMGuestSW                             B.04.30        Integrity VM Guest Support Software
  VMKernelSW                            B.04.20        Integrity VM Kernel Software
  DHCPv4                                B.11.31        IPv4 Dynamic Host Configuration protocol daemons and utilities

### ignite2

#uname -a
HP-UX ignite2 B.11.23 U 9000/800 611309392 unlimited-user license

#swlist IGNITE
 IGNITE                C.7.7.98       HP-UX Installation Utilities (Ignite-UX)

[Martin Ilavsky · Accepted solution]

Well, it took me 12hrs to find the solution :/ .. But at least I have one.  As those tests above show there seems to be a problem that VM is connected to the same vswitch as the HPVM host. I don't know why. I did attach another NIC on HPVM server and connected it to intranet. I have deleted old switches and created new, dedicated switch:

#hpvmnet
Name     Number State   Mode      NamePPA  MAC Address    IPv4 Address
======== ====== ======= ========= ======== ============== ===============
localnet      1 Down    Shared             N/A            N/A
hpvmsw        5 Up      Shared    lan2     0x0017a476ad6d

#hpvmnet -Shpvmsw
Name     Number State   Mode      NamePPA  MAC Address    IPv4 Address
======== ====== ======= ========= ======== ============== ===============
hpvmsw        5 Up      Shared    lan2     0x0017a476ad6d

[Port Configuration Details]
Port    Port         Port     Untagged Number of    Active VM    Tagged
Number  State        Adaptor  VLANID   Reserved VMs              VLANIDs
======= ============ ======== ======== ============ ============ =============
1       Reserved     avio_lan none     1                         none
2       Reserved     avio_lan none     1                         none

Now everything is working like a charm. 

Question is: why? What caused discards of those packets during boot off? Because all VMs were able to communicate (ping) among themselfs when connected to the switch as shown before.

[Steven E. Protter]

Shalom,

You may not want to hear this but moving Ignite to an HPVM host puts another layer between the end user and Ignite server.

You got it to work, but you need to fully test all recovery scenarios.

I don't feel comfortable at this time putting an Ignite server in an HPVM host. I built a pair of Ignite servers in vpar systems on a superdome. That is so similar to doing a standalone Ignite server that I had no problems with it.

I did in that scenario to a complete test. I restored/built 11.11 PA-RISC 11.23 PA-RISC/IA-64 and 11.31 PA-RISC/IA-64 systems. I made sure I could restore systems into a HPVM guest as well.

That extra layer of networking having Ignite in an HPVM hosts makes me worry. I don't need that kind of complication if I'm trying to fix a broken system.

Note: If you disable legacy I/O before doing an Ignite backup on 11.23 or 11.31, a bug will not allow you to restore the system. This is a known problem with Ignite that I had to find out the hard way.

SEP

[Martin Ilavsky]

Hi,

yop, I know about the ignite bug with legacy DSFs :facepalm:

Yes, an extra layer - but only for VMs trying to do a recovery. Any other server communicates via lan0 - the interface which is not part of any hpvm switch. I did recovery tests with IA/PA on 11.23 and 11.31, physical and virtual - no problems yet. I didn't do any precise tests  but I haven't noticed any performance degradation during recovery on VMs either. 

I do have a backup igntie server if there is a need, so so far so good.

[Eric SAUBIGNAC]

Bonsoir,

Well, since you have found the solution, my post has no interest. It's a pitty that I din't found this thread earlier : we also had the problem ... and the same conclusion. I would have been able to avoid that you waste time. Sorry :smileysad:

Anyway, my post is just about a HPVM host as an ignite server. I don't see any kind of extra layer. It works pretty well, both for physical boxes and virtual ones. More, if you have several HPVM Hosts, you can configure multiple ignite servers on them. Then, backup guests from one HPVM host to an other, and vice versa. This a robust solution, particulary in DR scenarios.

Eric

[Martin Ilavsky]

Hi,

well .. too late ;) .. it's little bit chaotic as the ITRC is being migrated. I did search this new forum and didn't find anything useful. 

I think Steven meant logical layer, i.e. to have ignite working you have have to have HPVM working. So if you hit a bug with HPVM and all VMs go down, your igntie server goes down too.

The question is why it behaves as it does. But as it is not possible to run tcpdump on interface which is part of the vswitch, hard to say.