Ups and downs of performance

Staking & Delegation Operate a Stake Pool
1

Hey Cardanians!

Stakepool MESH I am operating has performance graph similar to bitcoin price over time. I have series of epochs that are far above 100% and some that are below 100%.

I operate on dedicated CPU vm’s on Linode that has 16GB ram for producer and 8 & 16 GB ram for 2 relays.

I am putting these words here because I love when everything goes smooth, putting my heart to this thing and have bad time when something goes wrong.

What I did so far that was unsuccessfully to bring my pool to work fully smoothly:

  • ugraded from shared to dedicated CPUs
  • keep KES keys rotated every couple of epochs
  • no swap
  • having nice number of relays in my topology that are checked are working
  • I am always up to date with cardano-node, now at 1.27.0 with Memtrace off
  • logs looks solid too, no nasty errors nor warnings

Average CPU usage in my producer is around 0.03 %, bit higher in relays.
I miss around 4-20 slots every 24 hours.

I do not use cntools this time.

What did you do to improve performance of your nodes?

2021-05-20_09-46
2021-05-20_09-461147×1075 236 KB

2

How do these figures compare to leaderlog predictions?

3

Hi @whihajste, from what I see your total luck counting from epoch 210 till now is 99% which actually confirms all is in order. This is just about luck swings that happen. In the end as long as the average is around 100% you are fine.

Of course if your nodes are down or you have some issues this will drop your success rate but otherwise its just luck and it will smooth out over time.

1 Like
4

leaderlog predictions ? How to determine/get it?

5

Like this Run Cardano Nodes - Cardano Docker

1 Like
6

How do you like Linode? Was it much more expensive to move to dedicated from shared?

7

I like it a lot, it is super reliable. Yeah, going from shared $80 per month to dedicated $120 per month hurts but I see it is much stable so I converted all my network to dedicated.

8

I’m super new to this, but from all the reading I have been doing I have seen people trace missed blocks to relays that are either very far from them, or ones that just timeout or drop connection a lot. I have been watching the basic console logging trying out different relays to see which ones are problematic. IOHK’s relays are pretty awful, I’ve found, for my server in Santa Clara. I use IP search tools to see where the relays really are as I have discovered the locations in that list are often false.

1 Like
9

I dream about the tool that runs on the relay and lists other relays based on the connection quality (measured at the relay it runs on).

2 Likes
10

Have you tried time synching with chrony ?

11

Yes, after a few days I can tell it seems to help nothing or just a small bit on my setup.