Hi,
I run a Ubuntu 10.04 LTS Desktop System 4.0 on a ESX server.
Ubuntu system:
4x Intel Xenon 2.13GHz
3 GiB RAM
Kernel: 2.6.32-30-generic
Now the problem:
The server is slow after some time (hours). The interface works just slow and rebooting also takes very long.
What could be due to?
Thanks in advance.
Hallo,
ich betreibe ein Ubuntu 10.04 LTS Desktop System auf einem 4.0 ESX Server.
Ubuntu System:
4x Intel Xenon 2,13 GHz
3 GiB RAM
Kernel: 2.6.32-30-generic
Jetzt zum Problem:
Der Server wird nach einiger Zeit (Stunden) langsam. Das Interface funktioniert nur noch schleppend und das Neustarten dauert ebenfalls
sehr lang.
Woran kann das den liegen?
Danke schonmal. 
Same issue, but with more supporting data 
.
I've been running Linux servers under ESX for years and have installed/managed >1000 vmware linux guests over the past few years. I run into this problem occasionally and have not yet found a solution.
Currently:
VMware ESXi 4.1.0 with vSphere4 essentials license, running on a:
Dell PowerEdge T610
2x Intel Xeon E5520 @ 2.27GHz (4 core)
8181MB ECC RAM
5x 7200RPM SATA discs in a raid5 configuration.
I have ~20 guests running on this server. 19 are running Ubuntu 10.04 server or 10.10 server with VMwareTools installed from VMware ESXi (not the hacky OpenVMtools).
With all the machine powered on and settled into their jobs for a few days, my CPU usage averages 1.4GHz over the course of 24 hours (at a resolution of 10 samples per minute). The RAM usage sits around 1.4GB active, 7GB consumed, and balloon is around 5GB. I'm aware this is cutting it a little close, but the symptoms I experience are present even when RAM consumption is under 10% (1.6GB consumed and 1.2GB active).
The problem is that while machines are super responsive at boot and remain so for a few hours, after a period of relative inactivity (databases are still serving out data, but no users are actively logged into the machine) the server performance comes to a crawl. Fetching a simple "hello world" in html from Apache takes >=10 seconds. The second request is instantaneous. Logging into a console (ssh), the machine takes 10 seconds to response to my first packet, presents me with a login prompt, I enter the username and wait for 10-15 seconds before prompted for the password, enter the password and wait 10-50 seconds before I'm finally logged in. I use ssh -vvvvvvvvvvv to make sure the problem isn't an ssh problem. The hang ups always occur while we're waiting for the server. Once logged in, I issue a simple 'ls' on a directory with 1 file and wait for 5-10 seconds for a response. Again, issuing the same command immediately, results in an immediate response.
This feels like swapping, but neither the VMguest nor the VMhost report any swapping. I've even logged into the VMhost via ssh to watch vmstat and according to vmstat, we're not waiting on IO. CPU load is negligable, RAM is fine, not waiting for IO, the server load is/should be very low.
The only way I've found to recover from this degraded performance is to reboot the VMguest. This does not seem to affect Windows guests, nor BSD variants. If I clone the affected machine to another VMhost, it runs fine. If I reinstall the VMhost the problem persists. It seems like either a hardware compat. issue or a problem with my configuration of the VMhost, but I've never been able to track it down, and have seen it occur in 20+ VMhosts.
The problem does not occur when I use another hypervisor technology such as KVM or even Virtualbox, on the same hardware.
Anyone else?