LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Networking (https://www.linuxquestions.org/questions/linux-networking-3/)
-   -   High amount of Packet Drops in Network Interface (https://www.linuxquestions.org/questions/linux-networking-3/high-amount-of-packet-drops-in-network-interface-4175466990/)

rajaniyer123 06-22-2013 09:20 AM

High amount of Packet Drops in Network Interface
 
Hi,

In one of my Media Server, we observed below issues.

bond0 has been configured with eth1 & eth2.

1. After some amount of RX Bytes approx 10 TB suddenly high amount of RX Drops observed which causes backup failures. After eth1 down/up issue resolved.

2. Also not same amount of traffic observed on both of the interfaces which are part of bonding.

3. In case of eth2 no RX Drops observed like eth1.


Please suggest in case you have not any solution.

Thanks
Rajan

Ser Olmy 06-23-2013 07:06 AM

Quote:

Originally Posted by rajaniyer123 (Post 4976574)
bond0 has been configured with eth1 & eth2.

1. After some amount of RX Bytes approx 10 TB suddenly high amount of RX Drops observed which causes backup failures. After eth1 down/up issue resolved.

This is probably hardware related. Do you get any messages in the logs when this happens?

The problem could be at the other end as well. What are these bonded interfaces connected to?

Quote:

Originally Posted by rajaniyer123 (Post 4976574)
2. Also not same amount of traffic observed on both of the interfaces which are part of bonding.

That is to be expected, especially for inbound traffic.

It all depends on the bonding policy. You can configure the bonding policy to be "round-robin", which means outbound traffic will be more or less equally distributed among the participating interfaces. However, you have no control over the bonding policy for incoming traffic, as that's decided by the equipment at the other end.

If your bonded interfaces are connected to another Linux box with a similar setup and a "round-robin" policy, you should expect incoming traffic to be just about equally distributed among the physical links. Switches, however, seldom support an RR policy, and will instead load balance based on MAC or IP addresses.

rajaniyer123 06-23-2013 07:31 AM

Hi
 
Hi,

Thanks for update.

This is probably hardware related. Do you get any messages in the logs when this happens? - No messages observed in log

The problem could be at the other end as well. What are these bonded interfaces connected to? - they are connected to switch

Thanks
Rajan

Ser Olmy 06-23-2013 07:54 AM

Quote:

Originally Posted by rajaniyer123 (Post 4977028)
No messages observed in log

None at all? Nothing in /var/log/messages? Nothing reported by dmesg? That is highly unusual.

What does ifconfig eth1 and ethtool eth1 report when these problems start occurring? Have you tried locking the speed and duplex settings at both ends (both the switch port and the NIC)?

Also, what kind of NIC are we talking about here?
Quote:

Originally Posted by rajaniyer123 (Post 4977028)
they are connected to switch

What make and model?

How is the bonding/teaming configured on the switch?

The switch should have some kind of log, or at least you should be able to view port statistics. What does it say about the port connected to eth1?


All times are GMT -5. The time now is 09:19 AM.