| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 5247.1 | | EEMELI::MOSER | Orienteers do it in the bush... | Mon Mar 10 1997 11:28 | 13 |
| the SDA output shows a state of 'online' and not 'unavailable' as you
mentioned in your title. 'Unavailable' is the state after the MC spur
has passed all the hardware init, but before there is a link ok to
either a real MC hub or the other node in a virtual hub is also up and
running.
In your case, there are currently 2 nodes online in a real hub config
with node id's 5 and 7.
This all looks ok, so where's the problem? or show us the same SDA
output from the other node.
/cmos
|
| 5247.2 | Confused? | TMCUKA::ROWELL | Paul Rowell @BBP | Mon Mar 10 1997 12:27 | 10 |
| The unavailable came from:
SHOW DEV MC
Device Device Error
Name Status Count
MCA0: Unavailable 0
May be I'm confused?
-Paul
|
| 5247.3 | No Systems Programming Interface yet! | EVMS::PERCIVAL | OpenVMS Cluster Engineering | Mon Mar 10 1997 17:20 | 8 |
| This is the usual behaviour with Memory Channel devices - SHOW will
always give a device status of Unavailable. You can only see the real
state through SDA.
I've asked the memory channel people and they say this is to prevent
people using the SPI which is not released as yet...
Hope this clears the mystery up a little!
|
| 5247.4 | Confused re Cluster_config also. | TMCUKA::ROWELL | Paul Rowell @BBP | Tue Mar 11 1997 03:43 | 16 |
| The other thing that added to my confusion was that when I tried to use
cluster_config to disable the LAN for comms it said:
WARNING: This node has no CI, DSSI, or Memory Channel enabled for
cluster
communications. Disabling the LAN will leave this node
unable
to participate in a cluster.
The configuration procedure has completed successfully.
This made me think that the MC wasn't working, so I then did the SHOW
DEV....
-Paul
|
| 5247.5 | | EEMELI::MOSER | Orienteers do it in the bush... | Tue Mar 11 1997 09:03 | 9 |
| was that the initial cluster_config run when PMDRIVER wasn't loaded?
$ SHOW CLUSTER /CONT
ADD LPORT
Do you see a PMA0 besides the PEA0? If not, then PMdriver is not loaded
and no SCS traffic going over MC. If there, then things are fine.
/cmos
|
| 5247.6 | A bug? | TMCUKA::ROWELL | Paul Rowell @BBP | Tue Mar 11 1997 13:02 | 27 |
| Currently
$SHOW CLUSTER/CONT
ADD LPORT
shows:
View of Cluster from system ID 64716 node: TMC2K4 11-MAR-1997
17:55:39
SYSTEMS MEMBERS CIRCUIT
NODE SOFTWARE STATUS LPORT
TMC2K4 VMS V7.1 MEMBER PMA0
PEA0
TMC2K1 VMS V7.1 MEMBER PMA0
PEA0
But if I try to disable LAN for cluster comms, from cluster_config then
I get the message:
WARNING: This node has no CI, DSSI, or Memory Channel enabled for
cluster communications. Disabling the LAN will leave this node unable
to participate in a cluster.
I reckon that's a bug!
-Paul
|
| 5247.7 | | EEMELI::MOSER | Orienteers do it in the bush... | Tue Mar 11 1997 13:30 | 8 |
| ok, that might be true that in CLUSTER_CONFIG and/or CLUSTER_CONFIG_LAN
is a bug.
You could use the LAVC$STOP_BUS in SYS$EXAMPLES to stop the cluster
traffic on the LAN adapter(s), and if your cluster stays up, you can
bet that MC works fine for you.
/cmos
|
| 5247.8 | | EEMELI::MOSER | Orienteers do it in the bush... | Tue Mar 11 1997 13:37 | 12 |
| you're right, both C_C and C_C_L use the following checks when checking
the interconnects:
lan_cluster - see if PEA0 exists
ci_cluster - see if PNA0 exists
dssi_cluster - see if PIA0 or PAA0 exists
mc_clust_this - here they should check for PMA0 existance, but it
looks like they do some strange things...
so please QAR this against CLUSTER_CONFIG and CLUSTER_CONFIG_LAN
/cmos
|
| 5247.9 | Here is the fix! | TMCUKA::ROWELL | Paul Rowell @BBP | Wed Mar 12 1997 06:20 | 24 |
| Yes I will log QAR.
The fix is as follows:
In the SUB CHECK_INTERCONNECT there is the following:
$IF host_arch .EQS. "Alpha"
$THEN
$ ci_cluster = F$GETDVI("PNA0","EXISTS")
$ dssi_cluster = F$GETDVI("PIA0","EXISTS") .OR. F$GETDVI("PAA0","EXISTS")
$ IF ci_cluster
$ THEN
$ IF F$GETDVI("PNA0","DEVTYPE") .EQ. 41
$ THEN
$ ci_cluster = "FALSE"
$ dssi_cluster = "TRUE"
$ ENDIF <--- There should be an extra ENDIF here
$ IF mc_config
$ THEN
$ GOSUB mc_check_if_enabled
$ ENDIF
$ ENDIF
$ENDIF <---- This ENDIF should be removed
|
| 5247.10 | | EVMS::MORONEY | vi vi vi - Editor of the Beast | Fri May 09 1997 16:22 | 6 |
| I put this fix in RAVEN, so you won't get that spurious message anymore once
it hits the streets.
Thanks for locating the problem.
-Mike
|