[pve-devel] quorum using Rasberry Pi as node

[Thomas Lamprecht]

On 05/23/2016 08:54 AM, Wolfgang Bumiller wrote:
>> On May 22, 2016 at 11:37 PM Thomas Lamprecht <t.lamprecht at proxmox.com> wrote:
>>
>>
>>
>>
>> On 22.05.2016 22:48, Michael Rasmussen wrote:
>>> On Sun, 22 May 2016 16:44:48 +0200 (CEST)
>>> Dietmar Maurer <dietmar at proxmox.com> wrote:
>>>
>>>>>> This would be an interesting task, but I would be more interested
>>>>>> in pve-cluster (to manage corosync.conf).
>>>>>>
>>> I have looked into this and this is what I have found out:
>>>
>>> $ file /usr/bin/create_pmxcfs_db
>>> /usr/bin/create_pmxcfs_db: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=160c6b57da02b9b0e86be3b8995d7097ec8f7558, stripped
>>> $ file /usr/bin/pmxcfs
>>> /usr/bin/pmxcfs: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=dd96ad8a35fee8a9013efcc17dd20aa0c69231bb, stripped
>>>
>>> Dependencies:
>>> libcorosync4-pve
>>> corosync-pve
>>> libpve-common-perl
>>>
>>> libcorosync4-pve and corosync-pve could be replaced with the versions from jessie-backports (2.3.5-3~bpo8+1)
>>>
>>> 1) To have this working a porting of create_pmxcfs_db and pmxcfs will
>>> have to be done to armhf.
>>> 2) pve-manager will need to be aware of ignore_nodes: <list_of_nodes>
>>> in datacenter.cfg. Possibly configurable from gui.
>>>
>>> Is there something I386 specific in create_pmxcfs_db and pmxcfs?
>>>
>> Raspberry 1 here with Raspian Jessie, did not care for build times and its good to test low performance ;)
>>
>> I built packages for libqb, corosync-pve  and pve-common, no big problems, only two qb test fail but only for test timeout reasons.
> Did you verify the syscall numbers in pve-common?

I'd did had time to check that yet, but good hint, thanks.

>
>> I'm onto pve-cluster, there are some portability issues, mostly things which would fixed by setting W-no-format.
>> It's stuff like size_t variable but %lu format and uint64_t but %zX, ... nothing all to problematic.
> For 'size' there's %z, but for fixed-size types like uint64_t you'll probably want to use the PRI* macros or cast the parameter (which is easier to read).

Yeah, already using them :) I only didn't found a macro for hex-printing
a guint64 there is only one for decimal printing (G_GUINT64_FORMAT),
will cast them in the meanwhile.