[DRBD-user] hang during writes?

Matthew Wilson matthew at sixapart.com
Wed Jun 1 01:51:26 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi All,

I'm new to DRBD, but getting things together pretty well, except for  
one strange problem.  First let me describe my setup:

A pair of dual Xeon systems, (32 bit) 2.4GHz w/ 8G RAM.  Disk is RAID  
10 (4 x 73G) on LSI MegaRAID.  The systems are connected to a couple  
of networks each (10.0. and 10.10.), but also have a direct connect  
via an Intel server NIC in one of the PCI slots (10.11.) for DRBD.   
They are called Mad and Cow.

The systems are running Debian Sarge and linux 2.6.10 from kernel.org  
with DRBD loaded as a module.

I sat down and read the docs, got DRBD built, installed and running  
with very little trouble.  I now have the drbd synced and formatted  
with XFS and mostly everything works fine.

Here's the problem.  For my initial testing I did some copies from  
the local disk to the drbd.  Specifically I copied my mysql innologs  
(1G of data, 3 files)  from the root fs to the drbd.  Initially  
everything seemed great, the copies took about 4 seconds which is  
absurdly fast.  I ran it a few times for more data points.  Only,  
about the 4th time, it took over a minute to copy.

I have since done more testing, using `rsync -avPW /innologs /data/ 
innologs` so I can watch the progress.  Normally rsync takes maybe  
6-10 seconds to complete the copy.  Again though sometimes the copies  
take much longer.  Using rsync I can see when the copy hangs and I  
looked at the two machines and could see that they machine doing the  
copy was pretty much idle.   I also grabbed the output of /proc/drbd  
for each during the hang:


mad:~# cat /proc/drbd
version: 0.7.10 (api:77/proto:74)
SVN Revision: 1743 build by phil at mescal, 2005-01-31 12:22:07
0: cs:Connected st:Primary/Secondary ld:Consistent
     ns:154976123 nr:0 dw:11566455 dr:143410114 al:2892 bm:8754 lo:2  
pe:63 ua:0 ap:62

cow:~# cat /proc/drbd
version: 0.7.10 (api:77/proto:74)
SVN Revision: 1743 build by phil at mescal, 2005-01-31 12:22:07
0: cs:Connected st:Secondary/Primary ld:Consistent
     ns:0 nr:155107165 dw:155107165 dr:0 al:0 bm:8754 lo:89 pe:0 ua: 
90 ap:0

Does anyone have any ideas as to where I might look for answers?  I  
can replicate fairly easily so just ask if there's any data you need  
to diagnose this.

Thanks,
Matthew



More information about the drbd-user mailing list