Routing OSPF and Network Team

After a server move to the failover Data Centre – a file copy between the two sites slowed down by 6x.  I ran the copy a few times to confirm the slow down.

The file is part of a SQL Server ETL process. The ETL process transforms the data and updates the Datawarehouse,commits sime  SQL Server statistics and Datawarehouse  and prepares for a daily reconciliation.

As a DBA it can be difficult to prove a process is slow due to network issues- particularly if the issue is not experienced by everyone.  The Network team can be intransigent . The Network is critical to overall systems performance and application owners must have confidence.

My approach in working with Networks is to supply as much evidence as possible, and then they can investigate issues.

 First step was to present some evidence to the Network Team . The traceroute one way was:

 Tracing route to server1.my.domain.net [10.140.57.27] over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  10.140.144.254

  2    <1 ms    <1 ms    <1 ms  10.140.34.21

  3    <1 ms    <1 ms    <1 ms  10.140.34.9

  4    <1 ms    <1 ms    <1 ms  10.140.57.253

  5    <1 ms     1 ms     1 ms  10.130.85.252

   6     2 ms     1 ms     2 ms  server1.my.domain.net [10.140.57.27]

Trace complete.

 The traceroute the other way was:

 Tracing route to server2.my.domain.net [10.140.150.36] over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  10.140.27.254

  2    <1 ms    <1 ms    <1 ms  10.122.34.30

  3     2 ms     2 ms     2 ms      10.140.34.65

  4     3 ms     1 ms     1 ms     10.140.34.69

  5     1 ms    <1 ms    <1 ms   10.140.34.22

   6     1 ms    <1 ms    <1 ms  server2.my.domain.net [10.140.150.36]

Trace complete.

 The problem is one way the traffic was routed in an efficient manner. The other way the traffic was routed inefficiently.

 Working with the Network team , we noticed the link-state database on physically adjacent routers had not been updated . We issued the show ip ospf neighbour command on the Cisco routers. The command revealed nothing.

After some investigation the Network team discovered the ospf on the neighbouring router had not been enabled. Using show ip ospf interface  verified  that ospf was not enabled.

Once the ospf was enabled on the neighbouring router , the link updates finished and the file copy returned to the usual faster duration

A good find – based on  some  evidence and teamwork!

Related Posts

How to report a SQL Server performance problem

 

Author: Jack Vamvas (http://www.sqlserver-dba.com)

Nic Teaming

Network card teaming or NIC teaming are networking terms describing methods of combining multiple parallel network connections . The main purposes are 1) increasing throughput beyond a single connection and 2)  redundancy.

 The three main teaming configurations

 Network Fault Tolerance (NFT)

1)  Two – Eight Physical NICS are teamed together into one virtual network adapter.

2)  Only one port (Primary) is used for transmit and receive. The remainder adapters are deemed secondary Ports. They are used in the event of a Primary Port failure.

3) Heartbeats can be received and transmitted by all ports

Transmit Load Balancing with Fault Tolerance

1) All NFT features plus Transmit Load Balancing  (TLB)

2) Two – Eight Physical NICS are teamed together into one virtual network adapter

3) TLB allows   allows load balancing on IP traffic tramsmitted from server

4) Incoming traffic is not load balanced i.e Primary Adapter  manages all incoming traffic

Switch Assisted Load Balancing (SLB) with Fault Tolerance

1)     All NFT and  TLB features plus load balancing incoming traffic

2)     Load balancing incoming and transmit traffic

3)     SLB load balances all traffic regardless of protocol

Related Posts

Speed and Duplex settings

Author: Jack Vamvas (http://www.sqlserver-dba.com)