7

We are experiencing occasional hangs with ubuntu 16.04 LTS (kernel 4.4.0-34-generic) and we wonder if somebody has some clues how to get to bottom of this.

The log looks like this:

Aug 24 10:02:04 Ubuntu-1404-trusty-64-minimal kernel: [ 2046.519099] perf    interrupt took too long (2505 > 2500), lowering kernel.perf_event_max_sample_rate to 50000
Aug 24 10:03:19 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 43 of user ubuntu.
Aug 24 10:06:21 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 44 of user ubuntu.
Aug 24 10:06:22 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 45 of user ubuntu.
Aug 24 10:08:14 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 46 of user ubuntu.
Aug 24 10:17:01 Ubuntu-1404-trusty-64-minimal CRON[150011]: (root) CMD (   cd / && run-parts --report /etc/cron.hourly)
Aug 24 10:20:11 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 48 of user ubuntu.
Aug 24 10:26:28 Ubuntu-1404-trusty-64-minimal systemd[1]: Started Session 49 of user ubuntu.
^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@

The problem only happens with machines with SSD disk attached.

The memory and disk is checked and 100% ok.

The problem happens only on ubuntu 16.04.

Ubuntu 14.04 LTS does not run into this problem.

So this is definitely some software bug. The questions here is: What actions one can take to troubleshoot this? Which logs / debugs we should enable to get to the bottom of this?

user2196351
  • 191
  • 3
  • 1
    I have witnessed this issue on several 12.04 Ubuntu machines over the years and have never found a conclusive answer. One affected machine did fail a memtest so my suspicion is it's memory related. – Mark Riddell Aug 25 '16 at 18:10

1 Answers1

0

What file system are you running? Are journaling data or just metadata? It might be that the log file size (metadata) is extended and the change is journaled, but the log contents (data) isn't written yet. If the server then crashes or reboots the metadata is replayed from journal, but there is no data to recover so empty (zero) contents are shown. Do you use TRIM on your SSD? that would increase the likelihood of zero bytes (^@)

ptman
  • 27,124
  • 2
  • 26
  • 45