Droplet Performance after re-size

March 11, 2018 983 views
Server Optimization CentOS

Over the last few days I have resized (memory and drive space due to changes in the pricing structure) three droplets on NYC3, I have noticed a decrease in performance with all three but more so with our main webserver.

2 droplets original size 2GB RAM 40GB HDD 2 vCPU's changed to 4GB RAM 80GB HDD 2 vCPU's. One is the webserver the other provides database services.

1 droplet original size 1GB RAM 30GB HDD 1 vCPU's changed to2GB RAM 50GB HDD 1 vCPU's. This is a development droplet.

Prior to the resize the website (solarpanelstore.com) when checked with GTmetric had an approximate load time around 3 seconds now it is closer to 6+ seconds, looking at the waterfall graph I notice that there is a considerable long wait time before anything happens (greater than 4 secs). Also CPU usage has gone up dramatically compared to before the resize.

With the other two droplets I have noticed an increase in disk latency.

I had expected after the resize to benefit from the increase in RAM but at the moment that is not the case in fact I have had a performance decrease.

Currently the configuration for each droplet is the same pre and post resize therefore I would expect at least a status quo. (I can produce DataDog graphics which illustrate the changes).

Wondering if I need to change data centers to get better performances. Any help appreciated.

3 Answers

I would recommend opening a ticket with our support team sharing as much detail as possible. When increasing the size of a droplet our system will first check to see if the resources for the larger plan are available on the physical hypervisor that your droplet is on. If not, then the droplet is migrated to a new hypervisor.

While our team monitors all our hypervisors in order to proactively react to any performance issue this relies on automated systems to detect things like a user's droplet using massive amounts of CPU or Network so they can ensure optimal performance and address abusive uses of the platform.

My guess based solely on what you described is that your droplets may have been migrated to a new node that is seeing activity that may be just below the detection threshold for those automated systems. If this is the case, our team will be able to investigate and either address the problem on the existing hypervisor or assist in migration to a different node.

The support team will be able to quickly see if this is the case or not and it is entirely possible that they will discover a different reason for what you are seeing since they have much more information to go on and reaching out to them is the best way to see that this is addressed and resolved.

Thanks

I have already opened a ticket with support, we will see if your guess is the answer (I do believe you are correct because that's the conclusion I came to).
I will keep this post updated and hopefully post a solution.

  • Have you ever found the solution? I'm suffering from the same problem. My load time went from 0.5 seconds to 7 seconds.

I'm having this same issue. I resized a droplet and it is now totally unmanageable. I tried sizing it back down but it is still totally unmanageable.

Have another answer? Share your knowledge.