Mnesia and Oracle
Fri Aug 11 01:12:33 CEST 2006
No, IMHO as much as one may anticipate these kind of problems to be WAN
related (which is to be expected, and does happen often), they do occur on
LAN as well. Say, guys that are responsible for upgrading software on local
switch, and one networking company from San FranCISCO does that whenever
they feel like it, well these guys, for whatever reason, like to change the
settings for the port -- thus, auto_negotiate flag that was set to false,
end up being true, what used to be full duplex ends up being half... and as
the traffic volume picks up, so do the error/collisions etc. After a while
heartbeat message that is send to one node is going to be delayed enough (or
rather response to it might be buried under thousands of messages) to render
the connection to such a node invalid. When that happens, mnesia get
partitioned, eventhough connection to that node may be restored.
----- Original Message -----
From: "Joel Reymont" <joelr1@REDACTED>
To: "Valentin Micic" <valentin@REDACTED>
Cc: "Yariv Sadan" <yarivvv@REDACTED>; "Ryan Rawson" <ryanobjc@REDACTED>;
"Inswitch Solutions" <erlang@REDACTED>; <erlang-questions@REDACTED>
Sent: Thursday, August 10, 2006 11:39 PM
Subject: Re: Mnesia and Oracle
> On Aug 10, 2006, at 10:28 PM, Valentin Micic wrote:
>> Look, despite the fact that more and more money has been invested in the
>> network infrastructure, it seems that networks are less stable than
>> before. It is my experience that deploying distributed mnesia with table
>> replication over such a network, invariantly leads to problems -- mnesia
>> gets partitioned, and, depending on how complex the data model is, it
>> might be quite a difficult job to recover & re-syncronize.
> You are talking about a WAN as opposed to a LAN, right?
> There shouldn't be any replication problems in local gigabit network, for
> example, am I correct?
More information about the erlang-questions