Why multichain is using so much of Write IOPs

Question

Hi Team,

I 'm running few tests on multichain mainly to asses the performance. Whenever I put load on the node, the CPU gets spiked and so does my disk writes. What I 'm failing to understand is when I stop the load, the CPU usage goes back to normal (1-2%) my diskusage (write) remains very high, even after 30-40 mins after the load was stopped. So wanted to understand what is the writes being used for.

My size in mempool is one which should mean the blocks are confirmed.

 

27598                       -                    66572K                    147.5M                        0K                    100%                   multichaind

This is atop command. Similar WriteOp/sec I can see from AWS cloudwatch on my volume.

 

Thanks,

Ashish

Answer 1

We cannot know for sure without seeing logs, etc..., but it could well be background compaction of the database tables managed by LevelDB.

Comments

Even that was my 1st guess, background compaction kind of stuff or some indexing maybe. Is there any way by which I can confirm this? I mean any log file which I can check or send you guys to confirm.

---

With a normal commodity h/w when my node has more than 1million transactions, the disk io is constant at 50-60 iops even with 0 requests, I waited 1 hour after sending the last request, the iop/s didn't drop. That's concerning. Please help.