Andy, I want you to lean on this - I trust your gut feeling based on your xrootd experience... Jacek -------- Original Message -------- Subject: [JIRA] (DM-196) research non-blocking log4cxx appenders Date: Fri, 14 Mar 2014 14:56:41 -0700 From: Kian-Tat Lim (JIRA) <[log in to unmask]> To: Becla, Jacek <[log in to unmask]> Message Title Kian-Tat Lim <https://jira.lsstcorp.org/secure/ViewProfile.jspa?name=ktl> *commented* on an issue Re: research non-blocking log4cxx appenders <https://jira.lsstcorp.org/browse/DM-196> Do we need to log so much information in production that logging will be a significant contributor to (lack of) performance? For example, are we intending to log significantly more than MySQL? If we use asynchronous appenders, the logging may be unreliable (in addition to being substantially more complex). Is that a reasonable trade-off? If (synchronous) writing to files is sufficiently performant for Web servers and the like, why is it insufficient for Qserv? Add Comment <https://jira.lsstcorp.org/browse/DM-196#add-comment> Add Comment <https://jira.lsstcorp.org/browse/DM-196#add-comment> Data Management <https://jira.lsstcorp.org/browse/DM> / Improvement <https://jira.lsstcorp.org/browse/DM-196> DM-196 <https://jira.lsstcorp.org/browse/DM-196> research non-blocking log4cxx appenders <https://jira.lsstcorp.org/browse/DM-196> Default appenders in log4css are blocking (synchronous). This will dramatically impair performance of systems such as Qserv. We need to research the non-blocking appenders, and eventually switch to using them, in particular for production. This message was sent by Atlassian JIRA (v6.1.3#6158-sha1:b5b5eab) Atlassian logo ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the QSERV-L list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1