Closed superLiben closed 4 days ago
I conducted LTP stress testing on multiple services, and there were crashes and restarts. Please help analyze whether the problem is caused by an LTP bug. The version I am using is ltp20160510.
16 20:29:54 hgzmm3cs239 kernel: Call Trace:
Aug 16 20:29:54 hgzmm3cs239 kernel: [
17 06:27:24 hgzmm3cs239 kernel: Call Trace:
Aug 17 06:27:24 hgzmm3cs239 kernel: [
It seems memcg stress test kill the running test process. Since you used old ltp version , can you use the lastest ltp release?
看起来 memcg 压力测试杀死了正在运行的测试进程。由于您使用的是旧的 ltp 版本,您可以使用最新的 ltp 版本吗?
Thank you for your feedback. Our customer does not agree to change the version and needs to analyze the problem. So is there any way to prevent the program from running out of memory? Or you can set the range memory capacity
In addition, is the memory usage of ltp endless? Does the program automatically allocate the server's existing available memory?
Yes, ltp case will calucalte the needed allocate memory according to the server machine . I doubt whether there has a bug in ltp old version or your kernel version or cgroup version. So I think the simlpest way is to remove the memcg regression test and emcg_stress entry. For the way to set skipped case, you can use runltp -S option.
Yes, ltp case will calucalte the needed allocate memory according to the server machine . I doubt whether there has a bug in ltp old version or your kernel version or cgroup version. So I think the simlpest way is to remove the memcg regression test and emcg_stress entry. For the way to set skipped case, you can use runltp -S option.
Thank you. The customer is using the centos7.9 kernel, so I am not sure whether the information I pasted is the cause of the crash and restart. I will try to skip it.
Dear experts, are there any bug issues with ltp20160510 and 3.18.8-1168.el7 kernel cgroup?"
First of all there are testcases where processes are killed by OOM inside cgroups, because they test if a process that allocates too much memory is killed properly in that case.
Secondly LTP from 2016 is not supported. Only the latest stable LTP version is supported.
We conducted an LTP stress test on the server, and the server restarted or crashed. Can the pressure generated by LTP be used to complete the analysis of hardware problems?