Hi,

On Tue, May 20, 2008 at 10:57:44AM -0400, Audet, Jean-Michel wrote:
> There are no differences if a go with smaller packets.  In
> fact, it looks to go a little bit worst. 
> 
> I have some questions:  
>    1) Is heartbeat made to support transfer for large amount of data?   
>    2) Is there any heartbeat users that do that kind of transfer with success?

CIB.

>    3) Is there another solutions available (CIB, other) ?
> 
>    4) Is transferring 1 MB of data too large?  I am a little
>    bit surprise because for me, transferring 1 MB on a 100 Mb
>    or 1Gb link is not a big deal.

It's just a bit unusual, I'd say.

> I have changed my switch, Ethernet chips and still get the same
> results.  When I am crossing +- 700K of data, it starts to work
> very badly.  

No acks from the receiver? Some buffers filling up?

> When I enable debugging (debug 100), I get an incalculable qty
> of "Retransmitting Pkt".

Enabling debugging makes things worse? Strange.

> Anymore information input will be greatly appreciated.

CIB is the main user of heartbeat. At times it transfers large
configurations (XML). Perhaps you could take a look at that code.
Right now it lives at http://hg.clusterlabs.org/pacemaker. See
lib/crm/common/heartbeat.c for example.

HTH.

Thanks,

Dejan

> Jean-Michel Audet
> 
> -----Message d'origine-----
> De?: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] De la part de Dejan 
> Muhamedagic
> Envoy??: Thursday, May 15, 2008 7:13 AM
> ??: General Linux-HA mailing list
> Objet?: Re: [Linux-HA] Lots of Retransmitting pkt
> 
> Hi,
> 
> On Wed, May 14, 2008 at 12:17:04PM -0400, Audet, Jean-Michel wrote:
> > Yes, it's an application that we made to sync up data between
> > to application.  I do use the heartbeat API to sync up data.  I
> > have approximately 1 MB of data to sync up and I split the data
> > into +- 30 packets of 32K.
> 
> Why don't you go with smaller packets? Larger packets might be
> better if you have jumbo frames enabled, but that doesn't exist
> for 100Mbps networks.
> 
> > See below the statistics, I don't see any errors or packet drops.
> > 
> > Here is one of your statements:
> > So, it may be that your application is pushing the communication layer's 
> > limits. Or the network limits. Hard to say without looking at the network.
> > 
> > What do you mean about "communication layer's limits"?  Network limits?
> > Is sending 1Mbytes of data is big?  This would be very surprising.
> 
> It's not, of course. I was just thinking about sending such big
> packets back-to-back given that the MTU is much smaller.
> 
> > Architecture is very simple.  2 nodes connected to a switch (ZNYX ZX5000). 
> > 
> > 
> > eth3      Link encap:Ethernet  HWaddr 00:A0:A5:5E:08:F8  
> >           inet addr:192.168.0.149  Bcast:192.168.0.255  Mask:255.255.255.0
> >           inet6 addr: fe80::2a0:a5ff:fe5e:8f8/64 Scope:Link
> >           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
> >           RX packets:184419 errors:0 dropped:0 overruns:0 frame:0
> >           TX packets:614979 errors:0 dropped:0 overruns:0 carrier:0
> >           collisions:0 txqueuelen:1000 
> >           RX bytes:43421083 (41.4 MiB)  TX bytes:767939664 (732.3 MiB)
> >           Base address:0x5c00 Memory:f61e0000-f6200000 
> > 
> > eth3:0    Link encap:Ethernet  HWaddr 00:A0:A5:5E:08:F8  
> >           inet addr:192.168.0.100  Bcast:192.168.0.255  Mask:255.255.255.0
> >           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
> >           Base address:0x5c00 Memory:f61e0000-f6200000 
> > 
> > lo        Link encap:Local Loopback  
> >           inet addr:127.0.0.1  Mask:255.0.0.0
> >           inet6 addr: ::1/128 Scope:Host
> >           UP LOOPBACK RUNNING  MTU:16436  Metric:1
> >           RX packets:26 errors:0 dropped:0 overruns:0 frame:0
> >           TX packets:26 errors:0 dropped:0 overruns:0 carrier:0
> >           collisions:0 txqueuelen:0 
> >           RX bytes:1820 (1.7 KiB)  TX bytes:1820 (1.7 KiB)
> > 
> > 
> > Anymore details will be greatly appreciated!
> 
> The only thing I can think of is to enable debugging and open a
> bugzilla. Please don't forget to include all relevant
> information.
> 
> Thanks,
> 
> Dejan
> 
> > Jean-Michel Audet
> > 
> > Concepteur Logiciel
> > Software Designer
> > Kontron Canada 
> > Tel: (450) 979-5320 
> > 1 800 354-4223 
> > Fax: (450) 437-8053 
> > <mailto:[EMAIL PROTECTED]> 
> > Sales:<http://www.kontron.com/> 
> > Support:<ftp://ftp.kontron.ca/Support/> 
> > 
> > L'information contenue dans le pr?sent document est la propri?t? de Kontron 
> > Canada Inc. et est divulgu?e en toute confidentialit?. Cette information ne 
> > doit pas ?tre r?vis?e, utilis?e, divulgu?e ou distribu?e sans le 
> > consentement ?crit explicite de Kontron Canada Inc. Si vous n'?tes pas le 
> > destinataire pr?vu et avez re?u cette communication par erreur, veuillez 
> > contacter l'originateur et supprimer toute copie.
> > 
> > The information contained in this document is confidential and property of 
> > Kontron Canada Inc. Any unauthorized review, use, disclosure or 
> > distribution is prohibited without express written consent of Kontron 
> > Canada Inc. If you are not the intended recipient, please contact the 
> > sender and destroy all copies of the original message and enclosed 
> > attachments.
> > 
> > 
> > -----Message d'origine-----
> > De?: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] De la part de Dejan 
> > Muhamedagic
> > Envoy??: Wednesday, May 14, 2008 11:16 AM
> > ??: General Linux-HA mailing list
> > Objet?: Re: [Linux-HA] Lots of Retransmitting pkt
> > 
> > Hi,
> > 
> > On Wed, May 14, 2008 at 08:58:13AM -0400, Audet, Jean-Michel wrote:
> > > Hi all, 
> > > 
> > >  
> > > 
> > > Hi, I have an implementation using heartbeat 2.1.2.  It is based on
> > > Master/Slave architecture.  I have only two nodes on my implementation.
> > > 
> > > 
> > >  
> > > 
> > > I have a lot of "Retransmitting pkt".  At the end, all the packets are
> > > received, but I think it is very slow.  My packets are all around
> > > 32Kbytes in size but on a 100MBits network not a lot of traffic, it
> > > should not be any problems.  I am using the function sendnodemsg to
> > > directly send message to my slave node.
> > 
> > You have a home-made application using the heartbeat API?
> > 
> > > Here is a trace of one of the message.  When sending for example 40
> > > packets, 10 to 20 will be a least retransmit once.  
> > > 
> > > May 13 14:49:43 node1 heartbeat: [29379]: info: Retransmitting pkt 92
> > > 
> > > May 13 14:49:43 node1 heartbeat: [29379]: info: msg size =32581,
> > > type=msg_type_xyz
> > > 
> > >  
> > > 
> > > Any idea on what can be the problem?  Need more informations?
> > 
> > Heartbeat uses UDP (unicast, broadcast, or multicast). Is you
> > network healthy? Did you check the interfaces statistics? Did you
> > try watching the traffic using the tcpdump (or similar) tool? Did
> > you try unicast if you're using broadcast?
> > 
> > I believe that often Heartbeat packets are close to or exceeding
> > 32k with more complex configurations or more than 3-4 nodes, but
> > very probably not very often, i.e. packets are big only when full
> > CIB is being sent. I don't think that there were many complaints
> > of similar nature. So, it may be that your application is pushing
> > the communication layer's limits. Or the network limits. Hard to
> > say without looking at the network.
> > 
> > Thanks,
> > 
> > Dejan
> > 
> > 
> > > Jean-Michel Audet
> > > Software, Kontron Canada
> > > 
> > >  
> > > 
> > > _______________________________________________
> > > Linux-HA mailing list
> > > [email protected]
> > > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > > See also: http://linux-ha.org/ReportingProblems
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to