We should check some information about OS at startup and provide warnings.

Examples of the typical gotchas:

• too low ulimit on number of files;
• too low ulimit on number of threads;
• too low free memory (2 GiB or less);
• memory overcommit is set to 2;
• CPU scaling governor is not "performance";
• Linux is not using fast "TSC" timer;
• low disk space in a volume with logs (less than 1 GiB);
• low disk space in a volume with data (less than 1 GiB);
• transparent huge pages are set to "always" (on old Linux kernels);
• disk readahead is disabled and it is rotational;
• RAID 5 or 6 is used with unsufficient stripe cache size;
• mdraid is being repaired or checked;
• the number of corrected ECC errors is high;
• there are thermal throttling events recently;
• TurboBoost is disabled;
• checksum of the loaded code segment does not correspond to some record in ELF file;
• none of the data volumes are on the largest mounted fs;
• data volume is using Docker overlay fs;
• too low value of /proc/sys/kernel/pid_max or /proc/sys/kernel/threads-max, should be more than 30000;
• too low system-wide limit on the number of opened files: Too many open files issue #25994 (comment)
• Linux kernel version is old and this version has known bugs;