Suggestion: Improve logging when /tmp is noexec

[samjones3]

Centos 7.5
ES 6.4.2
JVM 1.8, current as of October 26, 2018

I just spent some hrs chasing bootstrap issues that were all caused by /tmp being noexec

Can I suggest that a bootstrap test for this be explicitly added, so that it gets fixed FIRST?

Issues that /tmp as noexec caused (when I flipped the machine to non localhost):

bootstrap.memory_lock -- could not be set to true. Got all these errors when I set it to true, but the root issue was /tmp being noexec

system call filter -- bootstrap fail, found myself setting "bootstrap.system_call_filter: false" to get system running. It came up (well, came up further), but turns out... this bootstrap fail was an artifact of /tmp being noexec

Fixed /tmp, and I am rocking, with all bootstrap checks and no funny configs.

[elasticmachine]

Pinging @elastic/es-core-infra

[jasontedor]

Can I suggest that a bootstrap test for this be explicitly added, so that it gets fixed FIRST?

You can suggest it, but the bootstrap checks come after native code that relies on jna.tmpdir being executable executes.