Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Actually, that makes some sense. If the network is way faster than the sync,
and the dd (which doesn't even use the network) is the same speed, then
there is something bad wrong with the underlying device or configuration.
You are not going to be happy until you get that fixed. Then, the dd should
beat the sync by a good bit.
You can skip the dd and build your filesystem and use it RIGHT NOW, but
there is still something wrong and the performance will probably suck. But,
you can get some work done.
If fixing the underlying problem doesn't wipe your data, then an online
verify and disconnect/connect (of the secondary node) will get you synced up
with zero downtime.
hth
Dan
From: Matt Baer [mailto:mbaer at lrnet1.com]
Sent: Friday, August 05, 2011 12:34 PM
To: Dan Barker
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
Ok, looking at this, my guess is that dd and the sync would take roughly the
same amount of time. Actually, if I am to believe the output of the kill
command you included, using dd will actually take more time. So I just take
your previous instructions and omit the dd command to skip it?
I don't care if the sync takes forever, I just want to be able to DO
something while it's syncing.
On Fri, Aug 5, 2011 at 11:15 AM, Dan Barker <dbarker at visioncomm.net> wrote:
OK, You can skip the zero, but the devices won't pass an online verify. It
won't hurt anything. All the sectors that need to be synced will be synced.
The disks are mostly zeros, so an online verify wouldn't do much, but it's
nice to know.
To find out how long the DD will take, "kill -USR1 taskid" (frightening
command, but it makes dd tell you how far it's along and does NOT kill it.
The 1M blocksize will help a lot. dd defaults to 512).
If you do skip the dd, the first verify will identify all the non-zero
sectors, and they'll sync up 100's of times faster than a full sync.
Good Luck!
Dan
From: Matt Baer [mailto:mbaer at lrnet1.com]
Sent: Friday, August 05, 2011 12:10 PM
To: Dan Barker
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
Roger that. It's running, but they're 1.8TB a piece so it'll take a while.
Just wanted to let you know, no need for it to go to the list. Thanks for
the help thus far, it's been difficult to deal with this and I have to get
it running ASAP.
On Fri, Aug 5, 2011 at 11:06 AM, Dan Barker <dbarker at visioncomm.net> wrote:
That's not the backing device. The backing device is something like
/dev/sdb. The drbd device is called device in your config. The backing
device is called disk.
It should not be mounted.
Dan
From: Matt Baer [mailto:mbaer at lrnet1.com]
Sent: Friday, August 05, 2011 11:44 AM
To: Dan Barker
Cc: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
After I down all, it won't let me issue the dd command citing:
dd: opening `/dev/drbd0': Read-only file system
On Fri, Aug 5, 2011 at 10:37 AM, Dan Barker <dbarker at visioncomm.net> wrote:
We've been off-list for a few messages, but this is now interesting enough
to be public. I apologize for the top-posting. Please read backwards. Dan
Well, to be certain, I'd dd the disks to zeros individually, and then start
with them sync'd.
drbdadm down all
dd if=/dev/zero of=/drbdbackingdevice bs=1M oflag=direct
on both sides.
New Blank Disk:
===============
#On both nodes, initialize meta data and configure the device.
drbdadm -- --force create-md r0
#They need to do the initial handshake, so they know their sizes.
drbdadm up r0
#They are now Connected Secondary/Secondary Inconsistent/Inconsistent.
Generate a new current-uuid and clear the dirty bitmap.
drbdadm -- --clear-bitmap new-current-uuid r0
#They are now Connected Secondary/Secondary UpToDate/UpToDate.
drbdadm primary r0
Now, recreate your empty ext3 file system and you are in sync.
Dan
From: Matt Baer [mailto:mbaer at lrnet1.com]
Sent: Friday, August 05, 2011 11:26 AM
To: Dan Barker
Cc: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
Well we're the perfect supplements for each other because, as you can see, I
don't know DRBD from a hole in the wall.
Yes, it's a brand new resource. I don't think I would say it's full of
zeros, as it has a clean ext3 file system on it.
Yes, I would LOVE to skip the sync, I've been dealing with this for weeks
now and right when I was about to go live, I tested the failover and it
didn't work because of a service heartbeat wanted to start wasn't going all
that well. While troubleshooting, I lost my 100% perfectly live server and
have to start from scratch. Problem is I only have two days to do it and
the thing has to sync 1.8TB at 12MB/s. I have no idea where the bottleneck
could be. Two servers, a cable connecting eth1 to eth1, both are
auto-negotiated at 1gbps on their own /30 subnet. The only thing there
would be garbage NIC cards, possible, but not probable, or the cable, more
likely, but I've never had an issue with it until now. Freshly constructed
servers, too.
I tried the drbdsetup /dev/drbd0 syncer -r 120M, been running like that for
about 5 minutes now and it hasn't changed at all.
On Fri, Aug 5, 2011 at 10:13 AM, Dan Barker <dbarker at visioncomm.net> wrote:
Is this a brand new resource? Why are you doing a full sync? If it's brand
new (full of zeros), you can skip the sync. Instructions upon request.
Btw, I don't know why you are getting 12% of your requested syncer rate. I'm
not a hot-shot linux performace analyzer, but there is a bottleneck
somewhere. I get 25M routinely here on GB nics. I have my Syncer set to 25M.
It drops to about 14M (each) if 2 are syncing.
To change sync rate without stop/start drbd: drbdsetup /dev/drbd1 syncer -r
120M
AL Extents seems a bit low. I use 1801 (big prime number that felt about
right).
Dan
From: Matt Baer [mailto:mbaer at lrnet1.com]
Sent: Friday, August 05, 2011 11:03 AM
To: Dan Barker
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
Ok, revised /etc/drbd.conf and restarted DRBD with the following
common { syncer { rate 100M; al-extents 257; } }
And I'm getting from /proc/drbd:
GIT-hash: 3e69822d3bb4920a8c1bfdf7d647169eba7d2eb4 build by
buildsvn at c5-x8664-build, 2008-10-03 11:30:17
0: cs:SyncTarget st:Secondary/Primary ds:Inconsistent/UpToDate C r---
ns:0 nr:281056 dw:272864 dr:0 al:0 bm:16 lo:257 pe:1969 ua:256 ap:0
oos:1308428488
[>....................] sync'ed: 0.1% (1277762/1278028)M
finish: 25:57:39 speed: 13,904 (12,400) K/sec
And I only have one resource, r0. All it's syncing right now is the post
mkfs.ext3 /dev/drbd0
On Fri, Aug 5, 2011 at 9:51 AM, Dan Barker <dbarker at visioncomm.net> wrote:
You are still mixing megabits and megabytes. Your 1000 megabit pipe won't
take a 600 megabyte stream, or a 150 megabyte stream. The maximum is about
125 MBps.
DRBD talks (and is documented to talk) bytes. Most everyone else talks bits.
You don't mention the speed you are getting.
Also, if you have 3 resources syncing, each will try for the syncer limit.
So, to use 50% of your capacity to sync 3 resources, you'd specify the rate
as 21M. Note: you can change the rate on the fly, during a sync.
Dan
From: drbd-user-bounces at lists.linbit.com
[mailto:drbd-user-bounces at lists.linbit.com] On Behalf Of Matt Baer
Sent: Friday, August 05, 2011 8:51 AM
To: Caspar Smit
Cc: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] DRBD won't take 1G syncer rate
I was playing with the settings yesterday and it let me set it at 600M and
it didn't make a difference in the sync speed at all. I then tried it with
your suggestion, dropped it to 150M just to be safe. Still no difference.
I wonder what the deal is. Could it be that this is the initial sync?
On Aug 5, 2011 1:29 AM, "Caspar Smit" <c.smit at truebit.nl> wrote:
> Hi Matt,
>
> 1000M means 1000 Mb/s NOT 1000mbps. To reach 1000M you should have at
least
> one (probably two) 10gbit interface(s). Since you have two 1gbit
interfaces
> (bonded with balance-rr?) a value between 100M and around 170M would be
more
> appropiate.
>
> Kind regards,
> Caspar
> Op 5 aug. 2011 08:21 schreef "Matt Baer" <mbaer at lrnet1.com> het volgende:
>> When setting the syncer rate in drbd.conf to 1G, it won't start, citing
> that
>> 1G is invalid. Get the same thing with 1000M. Any clue as to why? It
>> explicitly states that <mbaer at lrnet1.com>1G is acceptable in the docs.
> I've
>> triple checked and both interfaces are auto-negotiated at 1000mbps full
>> duplex.
_______________________________________________
drbd-user mailing list
drbd-user at lists.linbit.com
http://lists.linbit.com/mailman/listinfo/drbd-user
--
Thanks,
Matt Baer
Network Manager
Leaf River Telephone Company
102 W Second Street
Leaf River, IL
815.738.2211 - Main Office
866.738.2211 - Toll Free
mbaer at lrnet1.com
--
Thanks,
Matt Baer
Network Manager
Leaf River Telephone Company
102 W Second Street
Leaf River, IL
815.738.2211 - Main Office
866.738.2211 - Toll Free
mbaer at lrnet1.com
--
Thanks,
Matt Baer
Network Manager
Leaf River Telephone Company
102 W Second Street
Leaf River, IL
815.738.2211 - Main Office
866.738.2211 - Toll Free
mbaer at lrnet1.com
--
Thanks,
Matt Baer
Network Manager
Leaf River Telephone Company
102 W Second Street
Leaf River, IL
815.738.2211 - Main Office
866.738.2211 - Toll Free
mbaer at lrnet1.com
--
Thanks,
Matt Baer
Network Manager
Leaf River Telephone Company
102 W Second Street
Leaf River, IL
815.738.2211 - Main Office
866.738.2211 - Toll Free
mbaer at lrnet1.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20110805/5b23c4fe/attachment.htm>