Page 1 of 1

Solaris 10 x86 guest performance degrades over time

Posted: 15. Jun 2021, 19:35
by VBoxRHEL7
I have a Solaris 10 x86 guest that seems to run fine at first then the performance degrades over about two to three weeks. By degrade I mean: pings to and from the Solaris VM slow down from .3 ms to 1 ms, intermittent delays when typing at the console, and in-house software utilizing TCP/IP sockets begins to experience read/write failures. The killer for me is that certain critical serial devices (Digi realport) suddenly become unusable without a system reboot (I can't just reload drivers and/or re-init the devices). After rebooting Solaris, the system runs again for another two to three weeks (I don't normally reboot the host OS or restart VirtualBox). This behavior used to occur every 4 to 5 days. I extended it to 2 to 3 weeks after I increased CPUs and enabled Nested VT-X/AMD-V and Nested Paging.

In comparison, we have other Solaris 10 x86 systems (not VMs) that are configured nearly identical to this one, same network environment, same hardware, same software, etc. They all run fine 365 days/year, 24x7, except for scheduled maintenance. I'm hoping someone might have previous experience with this behavior.

I can post a VM log file but I'm unsure when is the best time to capture the log.

My basic configuration is:
HPE Proliant DL360 64GB RAM, 4TB RAID HDD
Red Hat Enterprise Linux 7.9 Host
VirtualBox 6.1.18 w/Extension pack
Solaris 10 x86 Guest w/guest additions

Re: Solaris 10 x86 guest performance degrades over time

Posted: 15. Jun 2021, 19:52
by scottgus1
I certainly may not be able to help fix this, as I'm windows-centric and Linux is iffy to me, let alone Solaris. But Virtualbox keeps 4 vbox.logs for the VM. So if the last or previous run was of the 2-3-weeks slowdown, one of the logs should already show at least that run time if not the problem.

Zip the VM's whole Logs folder, and post the zip using the forum's Upload Attachment tab.

Meanwhile, take notes on when the next run's noticeable issues happen: date and time of day the VM was started, how many days since boot and what time of day an issue is discovered, etc. If the batch of logs you have now doesn't show anything, maybe a time-correlated log and list of events might help.