Mailing List Archive

DRBD change in ports?
Hi, folks! I'm back after an involuntary absence from system
administration, and ready to jump back into drbd. I know I'd promised
some documentation on installation, and that will be coming when this
all falls into place, but for now I need to get things working!

This morning I upgraded a pair of RedHat 6.1 servers from 0.5.3 to
0.5.5, and found that something broke for me along the way.

I have two drive partitions on two different drives that I wish to
mirror on my servers po1 and po2. When I start drbd and cat /proc/drbd,
I get the following:

version : 55

0: cs:WFConnection st:Secondary ns:0 nr:0 dw:0 dr:0 of:0
1: cs:Unconfigured st:Secondary ns:0 nr:0 dw:0 dr:0 of:0


Attempting to set up the 2nd partition manually, I get the following:
# drbdsetup /dev/nb1 /dev/hdb1 B po1 po2
ioctl() failed: Invalid argument
Local address(port) already in use.


Say what??? I can't see any reason for the port to be in use, but maybe I
didn't shut things down properly. I did a full reboot, but the same
thing happened - the second drive will not configure. I dropped back to
0.5.3 (and my hacked scripts) and the second drive mounted fine.

My drbd[0|1] files were unchanged in all of this, and the drbd1 file
does in fact indicate a different port to use than the default, which is
used in drbd0.

Any clues as to why I can't get the second device working in 0.5.5 are
greatly appreciated!


======================================================================
Joe Hartley - UNIX/network Consultant - jh@example.com
12 Emma G Lane, Narragansett, RI 02882 - vox 401.782.9042
Without deviation from the norm, "progress" is not possible. - FZappa
Re: DRBD change in ports? [ In reply to ]
Am Mon, 12 Jun 2000 schrieb Joe Hartley:
>Hi, folks! I'm back after an involuntary absence from system
>administration, and ready to jump back into drbd. I know I'd promised
>some documentation on installation, and that will be coming when this
>all falls into place, but for now I need to get things working!
>
>This morning I upgraded a pair of RedHat 6.1 servers from 0.5.3 to
>0.5.5, and found that something broke for me along the way.
>
>I have two drive partitions on two different drives that I wish to
>mirror on my servers po1 and po2. When I start drbd and cat /proc/drbd,
>I get the following:
>
>version : 55
>
>0: cs:WFConnection st:Secondary ns:0 nr:0 dw:0 dr:0 of:0
>1: cs:Unconfigured st:Secondary ns:0 nr:0 dw:0 dr:0 of:0
>
>
>Attempting to set up the 2nd partition manually, I get the following:
># drbdsetup /dev/nb1 /dev/hdb1 B po1 po2
>ioctl() failed: Invalid argument
>Local address(port) already in use.

Hi Joe,

Well, probabely you should try

drbdsetup /dev/nb1 /dev/hdb1 B po1:7789 po2:7789

If you omit the ":port", port 7788 is assumed.

>
>Say what??? I can't see any reason for the port to be in use, but maybe I
>didn't shut things down properly. I did a full reboot, but the same
>thing happened - the second drive will not configure. I dropped back to
>0.5.3 (and my hacked scripts) and the second drive mounted fine.
>
>My drbd[0|1] files were unchanged in all of this, and the drbd1 file
>does in fact indicate a different port to use than the default, which is
>used in drbd0.

Please send your drbd[0|1] files, maybe we can find the flaw :)

>Any clues as to why I can't get the second device working in 0.5.5 are
>greatly appreciated!
>

By the way, there is a related bug in 0.5.5. It will check the address (and
port) also against the addresses of unconfigured devices (which is nonsense).

This is fixed in CVS.

BTW, I am planing to release 0.5.6 next week...

-Philipp
Re: DRBD change in ports? [ In reply to ]
Hi Joe,

On Mon, 12 Jun 2000, Joe Hartley wrote:
) Say what??? I can't see any reason for the port to be in use, but maybe I
) didn't shut things down properly. I did a full reboot, but the same
) thing happened - the second drive will not configure. I dropped back to
) 0.5.3 (and my hacked scripts) and the second drive mounted fine.

Check if the drbd initscript is setting MASTER_IF and SLAVE_IF. If it is,
it is also probably not unsetting them just before the end of the loop, so
the new device will "inherit" the same variables and thus try to use the
same IP:port.

I fixed that in the CVS some time ago, I think I sent a message to the
list commenting it... :)

( Fábio Olivé Leite -* ConectivaLinux *- olive@example.com[.br] )
( PPGC/UFRGS MSc candidate -*- Advisor: Taisy Silva Weber )
( Linux - Distributed Systems - Fault Tolerance - Security - /etc )