Re: VxWorks VME Ctrl-X vs. Power Cycle

[Matt Rippa <mrippa@gemini.edu>]

Hi Han and Rod,

Thanks for all the valuable input.

I'll add that at Gemini we use a hybrid console server + power
control unit. They're installed on nearly all our IOC's.

-Matt

http://www.wti.com/p-170-rsm-8r8-1-console-power-hybrid-ethernet.aspx



On 03/24/2014 06:44 AM, Rod Nussbaumer wrote:
> At TRIUMF, we use the control system to power cycle some VME
> crates via PLC controlled relays. This uses EPICS access control
> to regulate who can operate the relay. There is also a sequence
> of deliberate user operations required, to prevent inadvertent
> power cycling.
>    ---   rod.
> 
> 
> 
> On 03/23/2014 06:29 PM, Jeong Han Lee wrote:
>> Hi Matt,
>>
>> Only to focus the 'power-cycling or rebooting' option, here I
>> want to
>> share my options as follows:
>>
>> * Serial server : if you can still access the CPU though
>> 'serial server'
>> at least (Digi Portserver, Moxa serial server and so on). If
>> you have
>> small number of VME crates, it is a reasonable solution.
>> However, it
>> does not work if VxWorks freezes completely.
>>
>> * VME Crate has an Ethernet access to activate 'Reset' crate. For
>> example, Wiener supports SNMPv1/v2c so far. So, simply snmp
>> command can
>> reset the Wiener Crate.
>>
>> * Remote Accessible Power Distribution Unit, e.g., APC AP7921, is
>> another option. It is independent upon the VME crate, and it
>> supports
>> SNMPv1/2c/3. So simply snmp command can reset a power outlet
>> which is
>> used for the VME crate.
>>
>> When I worked at JLab, I used all three and possible methods,
>> because
>> some VME crates, installed in a experimental hall, are limited
>> to access
>> due to operation procedures/time and radiation. By simple GUI
>> interface,
>> a operator or s shift crew could execute any 'reboot' or 'reset'
>> commands easily and it worked well. However, there was one
>> concern for
>> SNMPv1/2c security issue. At that time, I only the SNMP2c
>> within one
>> subnet and hided SNMP bare commands under GUI in order to
>> reduce any
>> security issue.
>>
>> Anyway, to resolve this issue, I asked the Wiener to support
>> SNMPv3 in
>> their VME crate controller firmware last year. And the Wiener is
>> developing "SNMPv3 support" now and may (I hope) deliver a
>> firmware that
>> supports SNMPv3, to our site this summer. We are currently
>> testing
>> response time of v1/2c/3 by using simple network
>> configurations. As you
>> know, v3 is extremely slow. Technically, ELMA has the very
>> early SNMPv3
>> (Read) support, but their support isn't good (Only Read, bad
>> MIB file
>> and so on). CAEN has no plan to support SNMPv3 yet.
>>
>> Thanks,
>> Han
>>
>>
>> On 03/22/2014 06:31 AM, Matt Rippa wrote:
>>> Thanks Steve.
>>>
>>> The better coding part is always good =). But my last hang-ups
>>> were testing a new vxworks kernel and using the cross-wind
>>> debugger. Yikes! I haven't done that in years.
>>>
>>> Sounds like better negotiating with my team is in order.
>>>
>>> -Matt
>>>
>>> On 03/21/2014 10:53 AM, Hartman, Steven M. wrote:
>>>>
>>>> On Mar 21, 2014, at 3:21 PM, Matt Rippa <mrippa@gemini.edu>
>>>> wrote:
>>>>
>>>>> Can a modified reboot handler help with any of these
>>>>> occurrences?
>>>>
>>>> With VxWorks, you can definitely get yourself into enough
>>>> trouble that ^X or reboot() won't be enough. A SYSRESET or
>>>> perhaps even a power cycle is needed. If you just have a
>>>> terminal shell, you are out of luck. SNS built an option for
>>>> sending a reset through the timing system to the timing
>>>> receiver board which issues a SYSRESET on the backplane
>>>> (although this is almost never used). Some VME crates have
>>>> their own network interface which can be used to send a reset
>>>> or power cycle. A network power cycler is another option. If
>>>> you don't have something like that, and the crate is far
>>>> away, you best option is careful coding.
>>>>
>>>
>>
>