Production postmortemThe random high CPU
time to read 2 min | 253 words

A customer complained that every now and then RavenDB is hitting 100% CPU and stays there. They were kind enough to provide a minidump, and I started the investigation.

I loaded the minidump to WinDB and started debugging. The first thing you do with high CPU is rung the “!runaway” command, which sorts the threads by how busy they are:

image

I switched to the first thread (39) and asked for its stack, I highlighted the interesting parts:

image

This is enough to have a strong suspicion on what is going on. I checked some of the other high CPU threads and my suspicion was confirmed, but even from this single stack trace it is enough.

Pretty much whenever you see a thread doing high CPU within the Dictionary class it means that you are accessing it in a concurrent manner. This is unsafe, and may lead to strange effects. One of them being an infinite loop.

In this case, several threads were caught in this infinite loop. The stack trace also told us where in RavenDB we are doing this, and from there we could confirm that indeed, there is a rare set of circumstances that can cause a timer to fire fast enough that the previous timer didn’t have a chance to complete, and both of these timers will modify the same dictionary, causing the issue.

Tweet Share Share 6 comments
Tags:

More posts in "Production postmortem" series:

  1. (24 Jul 2023) The dog ate my request
  2. (03 Jul 2023) ENOMEM when trying to free memory
  3. (27 Jan 2023) The server ate all my memory
  4. (23 Jan 2023) The big server that couldn’t handle the load
  5. (16 Jan 2023) The heisenbug server
  6. (03 Oct 2022) Do you trust this server?
  7. (15 Sep 2022) The missed indexing reference
  8. (05 Aug 2022) The allocating query
  9. (22 Jul 2022) Efficiency all the way to Out of Memory error
  10. (18 Jul 2022) Broken networks and compressed streams
  11. (13 Jul 2022) Your math is wrong, recursion doesn’t work this way
  12. (12 Jul 2022) The data corruption in the node.js stack
  13. (11 Jul 2022) Out of memory on a clear sky
  14. (29 Apr 2022) Deduplicating replication speed
  15. (25 Apr 2022) The network latency and the I/O spikes
  16. (22 Apr 2022) The encrypted database that was too big to replicate
  17. (20 Apr 2022) Misleading security and other production snafus
  18. (03 Jan 2022) An error on the first act will lead to data corruption on the second act…
  19. (13 Dec 2021) The memory leak that only happened on Linux
  20. (17 Sep 2021) The Guinness record for page faults & high CPU
  21. (07 Jan 2021) The file system limitation
  22. (23 Mar 2020) high CPU when there is little work to be done
  23. (21 Feb 2020) The self signed certificate that couldn’t
  24. (31 Jan 2020) The slow slowdown of large systems
  25. (07 Jun 2019) Printer out of paper and the RavenDB hang
  26. (18 Feb 2019) This data corruption bug requires 3 simultaneous race conditions
  27. (25 Dec 2018) Handled errors and the curse of recursive error handling
  28. (23 Nov 2018) The ARM is killing me
  29. (22 Feb 2018) The unavailable Linux server
  30. (06 Dec 2017) data corruption, a view from INSIDE the sausage
  31. (01 Dec 2017) The random high CPU
  32. (07 Aug 2017) 30% boost with a single line change
  33. (04 Aug 2017) The case of 99.99% percentile
  34. (02 Aug 2017) The lightly loaded trashing server
  35. (23 Aug 2016) The insidious cost of managed memory
  36. (05 Feb 2016) A null reference in our abstraction
  37. (27 Jan 2016) The Razor Suicide
  38. (13 Nov 2015) The case of the “it is slow on that machine (only)”
  39. (21 Oct 2015) The case of the slow index rebuild
  40. (22 Sep 2015) The case of the Unicode Poo
  41. (03 Sep 2015) The industry at large
  42. (01 Sep 2015) The case of the lying configuration file
  43. (31 Aug 2015) The case of the memory eater and high load
  44. (14 Aug 2015) The case of the man in the middle
  45. (05 Aug 2015) Reading the errors
  46. (29 Jul 2015) The evil licensing code
  47. (23 Jul 2015) The case of the native memory leak
  48. (16 Jul 2015) The case of the intransigent new database
  49. (13 Jul 2015) The case of the hung over server
  50. (09 Jul 2015) The case of the infected cluster