More mysterious CPU spikes with version 4.1.1

[Samuel Sciolla]

Hi ArchivesSpace community,

My organization is in the process of updating our ArchivesSpace instances to use version 4.1.1. On three separate testing instances we've now experienced spikes in CPU usage after days of low CPU usage levels. There does not seem to be any corresponding surge in network traffic, memory usage levels remain low, and I don't see any extra or unusual application or job activity occurring at the time of the spikes (admittedly I'm only logging up to INFO at this point). The frontend continues to respond more or less normally at the elevated levels. CPU usage returns to normal low levels once we restart the application server.

I've seen a couple other recent posts with similar observations. I figured I'd post here in case others have debugging ideas or similar experiences -- before I open a GitHub issue or contact ArchivesSpace support. Any help would be appreciated.

Thanks!

Sam Sciolla

University of Michigan Library

[Blake Carver]

From what I see it's almost always...

Bots/Traffic - Usually some kind of bot wants all your stuff and tries to crawl everything way too fast. Sometimes when the site slows to a crawl they will go away. There can be huge spikes in traffic that don't last long, so they can be easy to miss. If there happens to be a REALLY BIG resource a hit to that on the PUI can cause trouble as well.

Jobs - The spreadsheet importer can really slow things down. Other background jobs can do the same.

API - Many places have things that hit the API too fast and that'll slow it down.

Big/Complex records - Editing a resource with MANY children and MANY agents and/or subjects can really slow things down. This can be hard to pull out of the logs. A single hit to a monster resource on the PUI could cause trouble as well, loading that thing is a ton of hits.

Sometimes the spikes end up just being a mystery, but that's rare, and I don't think I've seen that happen more on 4.x any more often than any other version. 

---

From: archivesspac...@lyrasislists.org <archivesspac...@lyrasislists.org> on behalf of Samuel Sciolla <ssci...@umich.edu>
Sent: Monday, December 8, 2025 10:41 AM
To: Archivesspace_Users_Group <archivesspac...@lyrasislists.org>
Subject: [ArchivesSpace Users Group] More mysterious CPU spikes with version 4.1.1

 

You don't often get email from ssci...@umich.edu. Learn why this is important

--
You received this message because you are subscribed to the Google Groups "Archivesspace_Users_Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to Archivesspace_User...@lyrasislists.org.
To view this discussion visit https://groups.google.com/a/lyrasislists.org/d/msgid/Archivesspace_Users_Group/cf9144fd-6036-43a3-9370-20e1ed122a25n%40lyrasislists.org.

[Blake Carver]

 Just thought of one more thing to add to this list... there's probably other things I'm not thinking of as well.

Under Powered Hardware - Sometimes it helps to just add more power and do some tuning. It could just be that a little more traffic, or some extra people working more than normal will just be too much for the setup. https://docs.archivesspace.org/provisioning/tuning/

Bots/Traffic - Usually some kind of bot wants all your stuff and tries to crawl everything way too fast. Sometimes when the site slows to a crawl they will go away. There can be huge spikes in traffic that don't last long, so they can be easy to miss. If there happens to be a REALLY BIG resource a hit to that on the PUI can cause trouble as well.

Jobs - The spreadsheet importer can really slow things down. Other background jobs can do the same.

API - Many places have things that hit the API too fast and that'll slow it down.

Big/Complex records - Editing a resource with MANY children and MANY agents and/or subjects can really slow things down. This can be hard to pull out of the logs. A single hit to a monster resource on the PUI could cause trouble as well, loading that thing is a ton of hits.

---

From: Blake Carver <blake....@lyrasis.org>
Sent: Monday, December 8, 2025 11:13 AM
To: Archivesspace_Users_Group <archivesspac...@lyrasislists.org>
Subject: Re: [ArchivesSpace Users Group] More mysterious CPU spikes with version 4.1.1

 

[Samuel Sciolla]

Hi Blake,

Thank you for the reply and weighing in on the problem! We have seen performance issues related to some of the high-usage situations you're describing in the past, but I'm not sure if they are at play here. We are seeing these spikes (I call them spikes but they really are a sharp increase and then sustained high CPU levels) on test instances that are restricted to certain networks and with the public interface turned off. The metrics we have that are related to network activity don't show any increases at the time of the spikes. We also don't see any evidence of background jobs running. In a couple of the cases, we don't have the API exposed at all. As far as I know, usage of these instances has been restricted to very minimal testing. In all cases actually, the spikes began during the weekend (not at exactly the same time) when I would expect application activity to be nonexistent.

Thank you for the point about tuning, I do think we could see some benefits from that, but we haven't seen the same CPU increase behavior (that I can recall) with version 3.4.1 that we've been running for a while.

Best wishes,

Sam

[Joshua D. Shaw]

Have you set the log level to debug to get the full list of entries? I think I'd try that next to help pinpoint what might be going on.

jds

---

From: archivesspac...@lyrasislists.org <archivesspac...@lyrasislists.org> on behalf of Samuel Sciolla <ssci...@umich.edu>

Sent: Wednesday, December 10, 2025 10:09 AM
To: Archivesspace_Users_Group <archivesspac...@lyrasislists.org>
Cc: Blake Carver <blake....@lyrasis.org>

Subject: Re: [ArchivesSpace Users Group] More mysterious CPU spikes with version 4.1.1

To view this discussion visit https://groups.google.com/a/lyrasislists.org/d/msgid/Archivesspace_Users_Group/79e306f2-cea8-4810-beda-90ce79d85952n%40lyrasislists.org.

[Samuel Sciolla]

Hi Joshua,

Thanks, I have since changed the log level to debug, but still waiting to see if we see another spike. I did get logs for one spike at the info level, but all I saw was the repeated Solr indexing round stuff.

Best,

Sam