[pve-devel] Possible Bug after Update PVE - PMG (VM) Disk not found

Detlef Bracker bracker at 1awww.com
Sun Jun 24 13:18:37 CEST 2018


More Infos after many tests and my cognition is: the *proxmox backup is 
absolute unsecure* and
can been the proxmox HA too - why a wrong MBR will then wrong on the 
other machines too!

First find the problem - I thing it can been a bug in VMs on Proxmox 5.x 
- yesterday after host reboot!
*The VM with the Proxmox-Mail-Gateway cant start (no bootable disk found)*

About that I have restore differented day-backups to test when the 
problems begins.
Every backup I have restore in a new VM (VM-ID) !

Restore from Backup the 19th works perfect, BUT this was without updates 
after the 19th and this
was the backup I have copied from the other host under 4.x to this host 
with proxmox 5.x !

*Restore from Backup 20th and all other backups after that date dont 
boot - Reason: No bootable disk found*

In noVNC console for the VM:

     Booting from Hard Disk...
     Boot failed: not a bootable disk

     then they test to boot from net and so on and beginns again and again

Ok what I can see in host:

     qm stutdown 134
     qm start 134
     qm status 134 --verbose

         balloon: 4294967296
         ballooninfo:
             max_mem: 4294967296
             actual: 4294967296
         blockstat:
             sata0:
                 idle_time_ns: 13065185000
                 wr_operations: 0
                 failed_rd_operations: 0
                 rd_bytes: 5120
                 wr_highest_offset: 0
                 flush_operations: 0
                 wr_merged: 0
                 failed_wr_operations: 0
                 wr_bytes: 0
                 failed_flush_operations: 0
                 rd_operations: 10
                 rd_total_time_ns: 1756662
                 wr_total_time_ns: 0
                 invalid_wr_operations: 0
                 timed_stats:
                 invalid_rd_operations: 0
                 rd_merged: 0
                 invalid_flush_operations: 0
                 flush_total_time_ns: 0
         cpus: 1
         disk: 0
             diskread: 5120
             diskwrite: 0
             maxdisk: 26843545600
             maxmem: 4294967296
             mem: 26101510
             name: mailgateway.mydomain.xxx
         netin: 157073
         netout: 0
         nics:
             tap134i0:
                 netout: 0
                 netin: 157073
         pid: 25716
         qmpstatus: running
         status: running
         template:
         uptime: 242

Via RepairDisk in the backup from the 20th - the first backup of the VM, 
where was after restore not usable:

GParted Disk-Information:

     Device Information:

     Model: ATA QEMU HARDDISK
     Serial: QM00005
     SIze: 25 GiB
     Path: /dev/sda

Partition table: unrecognized *<<<< Where has change this information?*
     Heads: 255
     Sectors: 63
     Cylinders: 3263
     Total sectors: 52428800
     Sector size. 512

     Partition: unallocated 25 GB

All Backups after the 19th has not an usabel harddisk when restored! But 
the VM was running until yesterday - ok, yes without boot!

OK, then I test the restored VM the scenaries of again updates and off 
on - shutdown on and so on - all works fine!

19th Mail-Gateway 5.0-71
Update - no Kernels was update - only libgcrypt20
shutdown
start - started normal
stop
start -

System-Information for the Proxmox-Gateway on host with proxmox 4:

Linux .... 4.13.16-3-pve #1 SMP PVE 4.13.16-49 (Tue, 22 May 2018 ....)

the same in the version of the 19th on the host with proxmox 5:

Linux .... 4.13.16-3-pve #1 SMP PVE 4.13.16-49 (Tue, 22 May 2018 ....)

I dont now, that on the VM other updates was between 19th until yesterday
installed?! But in my tests I get not more updates!

*The questions:*

A) What has destroyed the boot record of the VM of the Mail-Gateway?
Hardware-Failures on the Host cant been (SSD-RAID - without wareout - SMART
passed)
B) Why the backup was running perfect without see, that the boot record
is defective? - Why example when an admin dont boot the Host expl. 100 days
and then the backups only exists expl. 14 days, then the VM cant recover why
the backup save the bad VM-harddrives?!

*For this situation in moment I must say: The Proxmox Backup are not 
secure! *





More information about the pve-devel mailing list