EIGRP Queries and Stuck in Active

Lessons Discussion
61

Hello Thomas

Thanks so much for sharing this information. You are indeed correct. I went in and created a NetworkLessons note further describing this behavior of EIGRP queries.

Once again, thanks for your contribution!

Laz

62

Hello,

I have a question regarding the SIA process and what happens after it:
The topology is this : R1 ↔ R2 ↔ R3

Say R1 looses a route and is querying R2. R2 then doesn’t know about the network either and queries R3. R3 sends a reply to R2, but say there is congestion between R2 and R1 so therefore R2 cant send the reply towards R1. So after 1.5 minutes, R1 sends a SIA query and places the route into the SIA-state. If R2 sends an SIA reply, then the neighbor adjacency won’t be dropped. However, if R2 wouldn’t send an SIA reply, because there is still congestion or whatever going on, then the adjacency would be dropped.

My question is about the retry attempts EIGRP makes. For each reliable packet (packets that need to be acked) there will be 16 retry attempts. First question here…does this mean the SIA query process happens within this 16 retry attempts for query messages? Second question: Once an SIA query is sent out from R1, it doesn’t care anymore about the initial queries to be acked, is that correct?

If I need to explain more what I mean, feel free to ask me to. :smiley:

Kind regards,
Mirko

63

Hello Mirko

EIGRP uses RTP (Reliable Transport Protocol) to send its messages. RTP is a reliable protocol, and requires that messages sent are acknowledged using RTP’s reliability mechanisms. Within this mechanism is the 16 retries that you mention. This process runs concurrently and independently of the SIA process described in the lesson. In fact, the 16 retries operate at the Transport layer (i.e. within the RTP protocol) while the SIA process operates at the Application layer (i.e. that of EIGRP itself). Now the retries that are attempted by an EIGRP router using RTP will typically take place based on the retransmission timeout (RTO) used by RTP, which is dynamically calculated based on network conditions. By default the RTO is set to 200 milliseconds.

The SIA process works with another timer, which is the SIA timer, which has a value of 180 seconds (3 minutes). Typically, if a query is not responded to after half of that time (90 seconds), the SIA process kicks in and the SIA query is sent.

Because of the large difference between the RTO and the SIA timer, it is almost certain that the 16 retries have already been attempted well before any SIA query is sent. However, these processes are completely independent and operate on different layers of the OSI model.

From the explanation above, you can probably answer this question. Because the processes are independent, R1 will not care at all about the initial queries. At the level of EIGRP, it doesn’t care at all about those original queries, it only deals with the SIA query and the process from there on.

I hope this has been helpful!

Laz