If memory serves, just set the HEART_BEAT_TIMEOUT to a non-zero value.

The -R is a private API, so it shouldn't be used in this manner. :)

On 19 Nov 2010, at 12:30, Jeroen Janssen wrote:

> Hi,
> 
> On Windows and RHEL5 couchdb currently seems to run without the 'heart' 
> command.
> I'm trying to understand how to turn this on.
> 
> I was hoping it was an option in the init.d script configuration and I
> found the "RECURSED" option (-R) that seems to use HEART_COMMAND and
> HEART_BEAT_TIMEOUT.
> 
> However when I add -R to the COUCHDB_OPTIONS in /etc/sysconfig/couchdb
> then the init script seems to hang in "Starting couchdb".
> 
> Do the current scripts in couchdb repository support running with
> 'heart' and how do I correctly turn this on?
> 
> Thanks in advance,
> 
> Jeroen
> 
> On Thu, Nov 18, 2010 at 12:28 PM, Robert Newson <[email protected]> 
> wrote:
>> I know that the Windows version does not run 'heart' which is the bit
>> that restarts beam if it crashes. If that's true in RHEL too, then the
>> crashes can be explained fairly simply.
>> 
>> B.
>> 
>> On Thu, Nov 18, 2010 at 10:15 AM, Sebastian Cohnen
>> <[email protected]> wrote:
>>> short reply, inline:
>>> 
>>> On 18.11.2010, at 10:49, Jeroen Janssen wrote:
>>> 
>>>> Hello,
>>>> 
>>>> I am currently investigating the migration of data to couchdb and am
>>>> experiencing couchdb crashes that are not visible in the couchdb.log
>>>> file.
>>>> These couchdb crashes have occured on Windows (couchdb 1.0.1, erlang
>>>> 14A) and RHEL5 (couchdb 1.0.1, erlang R12B).
>>>> 
>>>> I understand it could be a bit difficult to answer, but I was
>>>> wondering what kind of errors would result in couchdb not logging an
>>>> error in its logfile?
>>> 
>>> out of memory errors are one example. couch (actually beam.smp) will crash 
>>> and couch will be restarted. I've seen this in production when replicating 
>>> or compacting databases with documents having lots of revisions (not sure 
>>> if this will help you though).
>>> 
>>>> 
>>>> For the moment I am trying to run with debug logging and hope that I
>>>> can get some better logging to show you (the problem seems to be
>>>> reasonable reproduceable, although it takes a while to get to it).
>>>> 
>>>> The problem I have might have to due with me 'frequently' calling
>>>> compaction on the database while also new documents are inserted.
>>>> I do the compaction because there is a problem with appending to 4G+
>>>> files on the Windows platform (in Erlang).
>>>> As part of the migration process, I will compact the database after a
>>>> certain amount of new documents have been added since last compaction.
>>>> This is to ensure that I can get to a situation where all current data
>>>> is actually inside couchdb (which unfortunately I have not reached
>>>> yet)
>>>> 
>>>> I initially thought the problem was maybe due to Erlang on Windows,
>>>> but since I have also had the issue on a RHEL5 machine, I am posting
>>>> here to see if it could be something in couchdb itself.
>>>> 
>>>> Let me know if I can do something specific to get more (usefull) logging.
>>>> Unfortunately, due to the data that's inside the database I can not
>>>> send the actual database files for analysis, but I will try if I can
>>>> create a testscript that mimics the migration of my data to couchdb.
>>>> 
>>>> Best regards,
>>>> 
>>>> Jeroen Janssen
>>> 
>>> 
>> 

Reply via email to