Runtime hot-backup system using two device servers

General Development
1

Hello,

In order to use Tango for mission critical system where single moments of failure is not allowed, hot-backup architecture needs to be considered.

On my mind, there possibily are two plans.

[quote]Plan A

  • Active Tango server keeps scanning sensors/devices.
  • Standby Tango server constantly updated status from active server.
  • If active Tango server failed, the standby Tango server takes over scanning sensors/devices from the failed one.
    Which means the standby role now changed to active automatically.
  • Tango clients always switching to the active Tango server transparently.
  • Commands from Tango clients always forwarded to active Tango server.[/quote]

[quote]Plan B

  • Primary Tang server and backup Tango server scanning sensors/devices at the same time.
    So there’s no need for the backup Tango server sync status from the primary server.
  • The Tango clients choosing primary initially dedicated explicitly.
  • If the primary failed, then client changed to backup immediately
    Which means the backup role now changed to primary automatically.
  • Command from Tango clients always forwarded to the primary server[/quote]

I know we can assign multiple databases at TANGO_HOST.
But it seems only for using two separate database servers when building device server initially.

Please guide me how to do this.

2

Hi Naver,
unfortunately none of your hypothesys are simply feasible with TANGO.
Plan A: please note that each TANGO device server has a unique device name within a
TANGO facility, so active and standby TANGO device servers need to be different.
This is not going to be transparent for clients.
Plan B: basically menas having two active control paths on a single equipment, that
may be possible (depending on the equipment) but tipically discouraged. Morover, any
synchronization for equipment access has to be done by hand as there is no provoded
mechanism in TANGO for such a requirement. Moreover, the same comment as Plan A
applies here too.
Short answer: it will be much simpler to design and develop a robust/reliable TANGO
device server than setup such redundancy.
Hope this helps,
Lorenzo