Re: [Quickfixj-users] What are the dangers of FileStoreSync=N

[<tom...@up...>]

Hello,

I've encoutered two system crashes when running a FIX module based on
QuickFIX/J and having FileStoreSync=N. Both instances were on virtual
servers that crashed or terminated unexpectedly. 

After these crashes, starting sessions with QuickFIX/J no longer worked.
In these cases, the file with extension ".session" in the filestore
directory was corrupt. I simply removed all state files pertaining to the
session and then starting sessions started working again.

I've described this on the QuickFIX/J user list (mail title= Server crash
leaves QuickFIX/J engine (1.5.3) in corrupt state for
FileStorePath=filestore).

I've since turned on FileStoreSync=Y but have not experienced any server
crashes since. We do not have performance requirements like you mention
though. It has indeed been documented to be much slower with FileStoreSync=Y
and I suppose on SAN it's probably much worse. Perhaps optimizing the
connection to the SAN might help? Or having the SAN be more performant? (I
have no experience with either).

Regards,
--
Tom Tempelaere

-----Original Message-----
From: Kuhn, Simon <Sim...@si...> 
Sent: Friday, 8 February 2019 15:53
To: qui...@li...
Subject: [Quickfixj-users] What are the dangers of FileStoreSync=N

QuickFIX/J Documentation: http://www.quickfixj.org/documentation/
QuickFIX/J Support: http://www.quickfixj.org/support/