What are your thoughts on why the nic card interface keeps failing?
 

The scenario is this, the RS servers are connected to each other to communicate via network cards via a small lan.

1 server is the web server, other is the db server. the issue is for whatever reason, the interface on the db server side dies once in a while. the problem is quickly solved temperarily by re initializing the interface. in our case, via linux, ifdown eth1, ifup eth1.

the network card has been swapped and that didnt solve the problem.

a hardware diognostic has been run, and that didnt find any problems.

what other suggests as to why the network card interface would fail?

Im not that tech savy, and thus this baffles me as to whats causing the failure.

:noob:

not sure if this will solve the problem but my suggestion is mayb the nic card can't take the load requests and mayb upgrading the nic card with more memory on it will work

^ would make sense. my old router would do something similar from torrenting, it would just jam up eventually from the amount of data going through. the card might not be designed for that kind of load.

other weak idea is, whats the cable running between them like?

yeah, checking the cable was the first thing i asked the host. cable tester tested the patch cable to be fine.

i didnt know network cards even had different levels or even memory! hmm, i should ask them about that.

bcrdukes suggested also possibly looking at the mysql load causing the nic interface to fail.

MySQL load is next to nothing, even when the server is at its peak. We also run Memcached from the DB server, and that barely has a load as well.

tried getting them to replace it with another?

^

did you guys notice, RS was starting to go down when we had the EPIC bradford thread? 11000 views - i'm sure once that thread dies down , everything will be A-OK!

what about the slot the nic card is in? if the os capable of handling the load without getting confused?

web server and db server are twins, there is actually a load balancer between the 2 but because we are using it as 1 web, 1 db, the load balancer is not needed.

both are running on linux redhat enterprise.

apache, mysql, php are all up to date.

and that is why it baffles me.

Problem with the switch port that the database server is connected to?

Do you guys own the switch, or does the colo provider?

Assuming it's a managed switch, get to the console and do a dump on it to see if there are errors showing up in the log relating to that port.

The rx dropped packets are too high. I have a feeling the transceiver is broken.

I moved the memcached server back to the webserver, so at least there will be much fewer packets going back and forth. This should help keep it from crashing every hour I hope

also scheduled for a full hardware scan on both servers, and swapping to yet another nic card on both servers tommorow at 8am.

Judging by the output of the "dmesg" it's probably and very likely a driver issue on the NIC. The chipset is either a VIA Rhine or a National Semiconductor and has been a known issue on the Linux kernel 2.6.x.x. There's a patch if you do a Google search. Another possible solution is to get a shorter Cat5e cable between the two servers but I doubt it.