There is an issue where a VE program causes a segmentation fault when you execute it at first time after you install RHEL8 or CentOS8 and SX-Aurora TSUBASA software into SX-Aurora TSUBASA.
This issue also occurs when if the number of huge pages is not sufficient.
$ ./a.out
Segmentation fault (core dumped)
We'd like to share its workaround.
If you have already confirmed that one of VE programs run successfully, this workaround is not required. If you have already set the huge pages and restarted services (or restarted the VH), this workaround is not required.
At first, please confirm the huge pages are properly set.
# sysctl vm.nr_hugepages
vm.nr_hugepages = 3840
The required number of huge pages are varies depending on the combination of the models and the installed software. Please refer to "4.11 Configuration of HugePages" section in "SX-Aurora TSUBASA Installation Guide". If the huge pages are not properly set, please set it following the instructions in the guide.
Then, please restart services.
# /opt/nec/ve/sbin/terminate-all-ve-services
# /opt/nec/ve/sbin/start-all-ve-services
Please confirm the status of the VEs is ONLINE with the vecmd command. It can take a few minutes until it becomes ONLINE. So, please wait for a short, If the status is “UNINITIALIZED” or “OFFLINE”.
# /opt/nec/ve/bin/vecmd state get
After these steps, the VE program will run successfully.
We will improve VEOS which is a part of SX-Aurora TSUBASA software to avoid this issue and release it.
| Posted by NEC admin (administrator) on 21 June 2022 at 02:53. |
|