-
Notifications
You must be signed in to change notification settings - Fork 132
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Error in Load Testing with MySQL in Firecracker Environment - Buffer Size Issue #1999
Comments
The issue causing the above error messages has been fixed in #2002. To test this fix, you can create a new image with Ops by specifying |
The same I have tried with postgres(Image - francescolavra/postgres:16.0) with the updated kernel image ~/.ops/659f598/kernel.img Got the below memory error and kernel got crashed.
|
The above log does not show a kernel crash, it's the postgres program that triggered an unhandled SEGV signal and therefore has been terminated by the kernel. Anyway, I tried load-testing postgres with HammerDB but I couldn't reproduce this issue.
|
First run the below build schema tcl file
After this run the below tcl file the issue will be reproduced
|
I found some issues in both the postgres:16.0 package and the Nanos kernel. The postgres issues have been fixed in a new package I uploaded (francescolavra/postgres:16.2), while the kernel issues have been fixed in 95d807a. With these fixes, the HammerDB load test completes successfully. |
I tried testing it with the nanos-kernel version 95d807a But I couldn't get it to start. When I tried starting the Firecracker VM with the kernel image 95d807a, I encountered a segmentation error. It seems there's an issue with the kernel image. I've tested with other kernel versions, and PostgreSQL starts fine with those.
|
I'm unable to reproduce the issue here. Does PostrgreSQL start fine if you run it with Ops (i.e. with |
The command ops pkg load works fine with the given configuration. However, when I used the same generated image (postgres) and kernel image (95d807a) with Firecracker, I encountered the error mentioned earlier.
Command for image creation : Kernel Given to Firecracker: |
It still works fine here. My Firecracker config file is as follows:
Then I run Firecracker as below:
What Firecracker command line are you using? And what are the contents of your Firecracker config file? |
FC Command : ConfigFile :
I think I found the issue ,
to your given conf and working fine
May I know why it is causing the issue |
Depending on the address of the initial pages allocated in setup_initmap(), the map_setup_2mbpages() function may not find an existing PDPT and or PDT in the page tables, and thus may need to use a new PDPT and/or a new PDT. However, the PDPT and PDT addresses passed to this function correspond to in-use pages, and as such cannot be reused and assigned to new PTEs. Moreover, when creating a new PTE, the map_setup_2mbpages() function is simply OR-ing the page address with the page flags, failing to set the USER flag and thereby preventing any pages referenced by this directory entry from being mapped for user space access; this may cause bogus segmentation fault signals to be delivered to the user process (#1999 (comment)). This change fixes the above issues by amending map_setup_2mbpages() so that it takes a physical memory region from which to allocate any new pages it may need, and calls new_level_pte() when creating a new PTE.
OK, thanks, you uncovered another bug in the kernel. This is fixed in #2019, and you can retrieve the fixed kernel build with |
I'm facing an issue while conducting load test with HammerDB on MySQL within a Firecracker environment.
I'm encountering error messages related to buffer sizes and frame handling. Below is a snippet of the error messages:
Test Configurations:
HammerDB Testing Setup:
The text was updated successfully, but these errors were encountered: