Background of the issue:
We run a multilingual WordPress site with WPML. Right after updating WPML (core + String Translation + Media) to 4.8.1, our total server load roughly doubled. PHP-FPM and MariaDB CPU spikes appear during normal page views and especially when switching languages or rendering pages with many translated strings.
Symptoms:
Overall CPU and DB usage approximately doubled compared to pre-update. TTFB is noticeably worse, and random requests stall while MySQL shows long queries.
Questions:
What could be causing the severe performance regression after updating WPML to 4.8.1?
How can we resolve the increased CPU and DB usage after the update?
thanks for getting back and providing details instructions.
1) I have accessed the website as instructed, but can you add some difference to staging site Title / page so I am sure I am accession staging, not production? Since they share same URL. I do not wish to cause any changes to your production website.
2) How can I see the performance issue, slow load and queries? Please share steps. I am trying to check on the homepage by switching languages, but I do not see any difference in load time etc.
1. You can identify the staging environment by the “Talkpal Staging” label with a black icon (see attached screenshot for reference). This should help you distinguish it from production.
2. Regarding the performance issue: I can run a controlled stress test from my side, which will simulate high load and push the staging server’s CPU usage close to 99%. Could you please let me know the level of load you would like us to apply in order to best reproduce the issue?
1) Yes, I can see that kind of icon, so we are good.
2) I am not sure if this would be the best procedure, since it would be expected that there is issue and slower response since you manually pushed CPU to 99%, and server is overload.
I would need somehow test and see that issue happens with specific steps and confirm issue is related to WPML, not some other plugin, code etc. This meaning happening only with WPML plugin or related WPML actions. Confirm there is difference issue only when WPML is active etc.
You mentioned in first reply "CPU spikes appear during normal page views and especially when switching languages or rendering pages with many translated strings.".
Please let me know, maybe I have misunderstood your idea.
To clarify, our main concern: the performance issues started only after we upgraded the WPML plugin. For the past month prior to the upgrade, there were no such problems.
As part of our troubleshooting, we also tested by deactivating all other plugins except WPML on the production website when the CPU load was at 99%. The high CPU usage did not go down, and the server continued to run under heavy load.
Additionally, I generated a spike report with a custom script that illustrates these CPU load patterns. You can review it here: hidden link
I completely understand your concern and I’ll be glad to assist further.
The main challenge here is that we need to be able to see and verify the issue directly in order to understand what’s happening and what might be causing it. That’s why I’m asking if there’s any way for me to reproduce or check it. This would allow me to confirm the problem, debug it, and, if necessary, escalate it to our 2nd tier or developers for deeper investigation in case it turns out to be a WPML-related issue.
At the moment, I don’t have a way to verify or replicate the problem.
It’s helpful that we have the report for further analysis, but if this was a one-time or random occurrence, many factors could have influenced it, and it wouldn’t be reliable for me to make assumptions based on that alone.
I hope this makes sense. Please let me know if there’s any way we can check or debug the issue further.
Got it, thank you for clarifying. Here’s the plan: I am currently running a controlled stress test that keeps the server load at around 60–70% CPU. Please see the attached screenshot. If I notice any abnormal spikes during this test, I will inform you immediately so we can capture the issue in real time.
Regarding debugging — from our observations, the issue does not seem tied to a specific action. What we’ve noticed is that the performance degradation started directly after upgrading WPML. I’ve attached another screenshot where you can see the CPU usage trend, with an arrow marking the time of the WPML upgrade for reference.
Let me know if this approach works for you, and I’ll keep you updated if we detect unusual spikes during the current test.
Yes, I can see the spike and that is concerning impact of WPML, definitely something we should further check.
Currently on your staging site, is the state same as the one you shared? I mean is WPML causing such large impact for your CPU usage? You can easily check by enabling/disabling WPML plugins and compare différance. If yes, then that is good starting point and for us to proceed further.
We started a stress test with CPU load at around 70–75%.
After deactivating three plugins — WPML Multilingual CMS, WPML SEO, and WPML String Translation — the CPU load immediately dropped from 75% to 50%, and we also observed a significant decrease in disk read/write activity (see attached screenshot).
Another important observation: during testing, disk usage on the server (160 GB total) increased from 45% to 92% and continued rising. Once the WPML plugins were deactivated, disk usage dropped back down to 44%.
On production, we are still experiencing the issue. For now, we have increased CPU resources from 32 cores to 48 cores and are monitoring performance while we wait for further analysis.
You mean this one? (see attached screenshot)
I just run stress test and can't see any difference.
Stress test is running right now, and I will keep it next 5–6 hours, if it's help to you
Did you install this plugin? (see screenshot)
I think you did it on Production,
Also, I see "String translation > Auto register string " disabled on Production
