Mailing List Archive: timouts, retrys, and no connections, oh my.

timouts, retrys, and no connections, oh my.

May 24, 2000, 3:02 PM

Post #1 of 3 (1863 views)

Having been playing around with settings for the past 2 days, i decided to
rm everything i have done, and download the newest version from the CVS
tree.
That being said, i decided not to play with anymore options, i did a basic
insmod drdb.o, and
drbdsetup /dev/nb0 /dev/hda7 B frontal temporal

and

drbdsetup /dev/nb0 /dev/hda7 B temporal frontal

with almost the same results as i was getting with the seconds NIC's i had
configured in earlier experiements

list of problems i seem to be having:

the defualt replication setting, (256k?) seems to KILL my poor 10mb hub, the
collision light is solid. no big deal.. but after i get a
drbd: timeout detected! (pid=448)
drbd: sock_sendmsg returned 1568
the connections go back to a unconfigured state, and is very difficult to
cooerce them to talk again, it would be nice if it had some sort of retry
here.
or if the connections decided to return..

long story short: turbo-linux-cluster server kernel 2.2.12 on both
and i have to leave, locking up shop, will provide more detail tomarrow.

Re: timouts, retrys, and no connections, oh my. [ In reply to ]

philipp at example

May 24, 2000, 11:10 PM

Post #2 of 3 (1782 views)

Permalink

Am Don, 25 Mai 2000 schrieb timh@example.com:
>
>Having been playing around with settings for the past 2 days, i decided to
>rm everything i have done, and download the newest version from the CVS
>tree.
>That being said, i decided not to play with anymore options, i did a basic
>insmod drdb.o, and
>drbdsetup /dev/nb0 /dev/hda7 B frontal temporal
>
>and
>
>drbdsetup /dev/nb0 /dev/hda7 B temporal frontal
>
>with almost the same results as i was getting with the seconds NIC's i had
>configured in earlier experiements
>
>list of problems i seem to be having:
>
>the defualt replication setting, (256k?) seems to KILL my poor 10mb hub, the
>collision light is solid. no big deal.. but after i get a

After setting it up with the commands above, there should be virtually
no traffic caused by drbd.

The resynchronisation traffic (250KB/sec) will only occure, if you switch
one of them into Primary mode AND issue drbdsetup /dev/nb0 REPL there.

During "normal operation" thus you are accessing the /dev/nb0 device on the
primary node, drbd will use as much bandwith as it needs (and gets).

>drbd: timeout detected! (pid=448)
>drbd: sock_sendmsg returned 1568
>the connections go back to a unconfigured state, and is very difficult to
>cooerce them to talk again, it would be nice if it had some sort of retry
>here.
>or if the connections decided to return..

Well, the latest sources out of CVS already have this retry mechanism.
(But you know that, code in CVS is usually even less tested than the releases?)

From my /var/log/messages:
May 23 13:57:49 alf kernel: drbd: retrying to connect(pid=1432)
May 23 13:58:28 alf last message repeated 3 times

But, I think you have some sort of network problem.
I think you should find out first, why there are so manny collisions on your
lan.

-Philipp

Re: timouts, retrys, and no connections, oh my. [ In reply to ]

thomas.mangin at example

May 25, 2000, 1:41 AM

Post #3 of 3 (1806 views)

Permalink

> >drbd: timeout detected! (pid=448)
> >drbd: sock_sendmsg returned 1568
> >the connections go back to a unconfigured state, and is very difficult to
> >cooerce them to talk again, it would be nice if it had some sort of retry
> >here.
> >or if the connections decided to return..

I got the same problem and I never succeed to reconnect the two nodes

Thomas