Open openhpi2 opened 12 years ago
openhpid verbose openhpid_debug.txt.tar.gz
Original comment by: snifferdev
I suspect none used several snmp_bc handlers simultaneously before. May be handlers have some resource/variable/connection conflict.
Hard to say how to resolve the issue. We have no active IBM guys now in OpenHPI development. And no access to the hardware.
Original comment by: avpak
You are most probably right.
Isn't there a way to easily simulate this setup? I saw you've implemented automated tests, isn't this the way to go?
I won't be able to give you access to the hardware. If there is a quick and easy way I can debug this, I might give it a try by the end of the week.
Original comment by: snifferdev
Not sure I got the simulation proposal. Is there a way to make virtual blade server hardware?
Original comment by: avpak
Well I myself can't imagine there is, but for any protocol there is a way to setup a dummy application server, exposing only the basic functionality you need. For the sake of this bug, snmp server which should: accept connection, authenticate (not necesairly), identify chassis, query. Thats was my idea of simulation. It should be easy to setup a server, see: http://pysnmp.sourceforge.net/examples/2.x/index.html
This bug does not concern specific values returned by the chassis, but rather a general problem in the connection handling on ohpi side, so this kind of setup may suffice.
I thought you might have used such solutions. I know its sometimes hard to get instantly access to hardware.
As for the debugging on my side, can I provide more info on helping to solve that?
Original comment by: snifferdev
Well, I've never seen IBM hardware (save for hard drives and PC XT/AT) and have vague knowledge about SNMP. Can try but cannot say how much time it will get and if the result will be successful. As I've said there is no IBM guys around now.
Original comment by: avpak
I see. OK, let me know if there is anything more you need or any progress is made with this bug.
Original comment by: snifferdev
Original comment by: avpak
Original comment by: avpak
Hi,
We have couple IBM BC H chassises filled with blade servers. Today we tried running OpenHPI v 3.0.0, but we receive the same errors as with 2.14.1.
When OHPI is configured with only one chassis, then it works properly. However when we add additional ones (additional \'handler libsnmp_bc { ... } \' sections) the daemon/library wont communicate with those chassises.
Sometimes it would return data from only one chassis, but after a very long time.
All the chassises work fine when OHPI is configured with one handler configuration for the libsnmp_bc plugin.
[root@lap ~]# openhpid -n -c /etc/openhpi/openhpi.conf openhpid: CRIT: plugin.c:593: Cannot propagate auto-insert timeout to handler. openhpid: CRIT: plugin.c:593: Cannot propagate auto-insert timeout to handler. No log handling enabled - turning on stderr logging snmpget: Timeout snmpget: Timeout snmpget: Timeout snmp_bc: CRIT: snmp_bc_discover.c:105: Discovery failed. Error=BUSY.
See attached openhpid_debug.txt for more verbosity.
Reported by: snifferdev