So, there’s a WordPress site that I help manage and we noticed a strange network throughput issue when we were migrating it from a datacenter to AWS. The database isn’t that big, so I originally migrated the WP server into AWS and had it point back to the server at the datacenter hosting the DB.
This is when the fun started – for some reason the WP server is consuming a TON of data connecting to the DB server. This, obviously, was noticed by looking at the AWS bill!! 😉
After we promptly got the DB into an RDS Instance, the network usage is still astronomical, but at least it’s all “in house” and we don’t get billed by AWS. I did verify that the connections are just between the two servers and there wasn’t what appeared to be any crawlers/bots hammering it. Furthermore, looking at the Linux network stats from the old WP server, I could see that it was also doing this before the migration.
Trying to search the web for things like: WordPress high network utilization database server did not really yield anything too helpful. And as this is an “Amazon Linux 2 LTS Arm64 AMI” Instance, it seems that I’m rather limited with some of the Networking Tools which may be helpful in getting a good look at the data.
As I was writing this, I also started to look at phpMyAdmin. Originally I was under the impression that the Network Inbound for WP was not in line with the outbound from the DB. But WP, in the 10 days since the last reboot, it has 2.4TB of received data. Stats from phpMyAdmin show that in the 60 days the RDS server’s been online it has sent 17.5TB of data.
But this leads me to my two questions… 🙂
- Is this normal?
- If it’s not normal, then what can be done to rectify this? The DB is only 450MB, so just in the past 10 days that would mean that the entire database could have been transferred 5,333 times (or 22 times per hour!). The WP server isn’t backing up the DB (just using AWS RDS to manage backups) or connecting to the DB server with any methods other than WP.
I’ve run databases on different machines than the application for years (including some big Connections installs…) and have never seen this type of throughput happening between the systems at such a high rate (or at least have never noticed it, but I kept an eye on VMWare stats, etc…).
Just to confirm that we aren’t seeing anomalies between what’s coming in to WP and what’s going out of the DB server, the charts below are pretty clear that the data rate is consistent. 😉