Bug #12305: "can't remove str from str_id" at unregister_sym

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.3.0p0 (2015-12-25 revision 53290) [x86_64-linux]
• Backport: 2.1: UNKNOWN, 2.2: UNKNOWN, 2.3: UNKNOWN

[ks (Konstantin Starodubtsev)]

I'm experiencing random occurrences of "can't remove str from str_id" crash.

Message string looks like this one:

/srv/sim/.rvm/gems/ruby-2.3.0@env-manager/gems/mongo-2.2.4/lib/mongo/server/description.rb:508: [BUG] 0x007f4f400ab310 can't remove str from str_id (`..dO.)

The only constant thing in it is the "can't remove str from str_id" substring. Other things like address, source code location or symbol string value in brackets differ between crash reports (see attachments).

In a C level backtrace information the only common part is top of a trace:

-- C level backtrace information -------------------------------------------
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(rb_vm_bugreport+0x51f) [0x7f4f8b0106ff] vm_dump.c:688
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(rb_bug+0xcf) [0x7f4f8ae9a01f] error.c:420
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(unregister_sym.part.12+0x20) [0x7f4f8ae4aada] symbol.c:399
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(rb_gc_free_dsymbol+0x62) [0x7f4f8afa56a2] symbol.c:637
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(gc_sweep_step+0x621) [0x7f4f8aec0d01] gc.c:2211
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(heap_get_freeobj_from_next_freepage+0x72) [0x7f4f8ae498c3] gc.c:3624
/srv/sim/.rvm/rubies/ruby-2.3.0/lib/libruby.so.2.3(newobj_slowpath_wb_protected+0xf4) [0x7f4f8aec43d4] gc.c:1679

We hit this bug at least once per day, so should you need any other information, let me know.

Attached are three full [BUG] reports for three different occurrences.

[ks (Konstantin Starodubtsev)]

• Assignee set to core

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)
• Status changed from Open to Feedback

How to reproduce it?
Couldn't you show simplified/reduced script?

[ks (Konstantin Starodubtsev)]

Nobuyoshi Nakada wrote:

How to reproduce it?
Couldn't you show simplified/reduced script?

This is a web application running on top of padrino framework and it's using various gems to consume and process data from external systems.
Unfortunately, this bug happens only once or twice per day, and I couldn't find any pattern in bug occurrence. It may happen during virtually any user request. I couldn't reproduce bug on a test dataset, but it happens on a different hosts when running in production.

If you could point me to some information about how to pull more debug information from runtime, or increase probability of this bug to happen, or whatever else, I could try it and then attempt to make a script to reproduce the problem.

Thanks.

[shevegen (Robert A. Heiler)]

I think it is difficult to track it down if it only appears sometimes.

Perhaps you find some way to trigger it more often - if you don't see a pattern in the bug occurrence,
perhaps you can find a pattern in the ruby code that triggers it ultimately.

The C code for the bug should be here in the function unregister_sym():

https://github.com/brightbox/deb-ruby2.3/blob/master/symbol.c#L395

I am not sure who or what tries to unregister a symbol but I think you will
find some patterns in the coming days/weeks. Heisenbugs sometimes require
detective work.

[naruse (Yui NARUSE)]

• Assignee deleted (core)

[eramsey (Eric Ramsey)]

Was there any resolution to this issue? We are encountering the same issue now that we have upgraded our Ruby from 2.1.6 to 2.3.3. This happens randomly throughout our framework when we are trying to run regression.

Any and all information as to where to look and or how to fix would be greatly appreciated.

[jeremyevans0 (Jeremy Evans)]

• Status changed from Feedback to Closed