[rrd-users] lost of cache content when database in unavailable

[Steve HICKS]

Hello,

I have found a problem on 1.4.9 and I have wait to try with a more recent
version to verify if problem has been corrected, before sent this email.

On 1.5.5 the problem still present:

I have try to update an existing database using rrdcached, no problem
that's works perfecttly.

If I start my rrdcached daemon an try to update a database which doesn't
exist on remote server, no problem : rrdcached inform me that database is
not present and reject my update, so this context is cleanly adressed.

BUT, if in am updating a databse known by rrdcached AND if this databse
become unavailable (loss of filesystem, etc..) THEN on last flush daemon
will fail with following error message (get in rrdcached server's log,
nothing appears on client side because for client there is no problem, the
remote daemon still to accept to store new updates request in his cache):

mqueue_thread_main: rrd_update_r
(/probtp_outils/portstat_db/datas/PLPORTR1/datas/test.rrd) failed with
status -1. (opening
'/probtp_outils/portstat_db/datas/PLPORTR1/datas/test.rrd': No such file or
directory)


The result is the lost of all the updates present in cache at this moment..
so a lot of informations if flushed are scheduled with a large interval..

(for info I have journalization active, and this does not change anything)

(tested with 1.4.9 and 1.5.5. rrdtool/rrdcached versions)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.oetiker.ch/pipermail/rrd-users/attachments/20151124/0e2489f2/attachment.html>