From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gregory Farnum Subject: Re: [NewStore]About PGLog Workload With RocksDB Date: Tue, 8 Sep 2015 16:47:46 +0100 Message-ID: References: Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Return-path: Received: from mail-wi0-f180.google.com ([209.85.212.180]:37449 "EHLO mail-wi0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753551AbbIHPrr (ORCPT ); Tue, 8 Sep 2015 11:47:47 -0400 Received: by wicfx3 with SMTP id fx3so120151235wic.0 for ; Tue, 08 Sep 2015 08:47:46 -0700 (PDT) In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Haomai Wang Cc: Sage Weil , "ceph-devel@vger.kernel.org" On Tue, Sep 8, 2015 at 3:12 PM, Gregory Farnum wrote: > On Tue, Sep 8, 2015 at 3:06 PM, Haomai Wang wrote: >> Hit "Send" by accident for previous mail. :-( >> >> some points about pglog: >> 1. short-alive but frequency(HIGH) > > Is this really true? The default length of the log is 1000 entries, > and most OSDs have ~100 PGs, so on a hard drive running at 80 > writes/second that's about 100000 seconds (~27 hours) before we delete > an entry. In reality most deployments aren't writing that > quickly....and if something goes wrong with the PG we increase to > 10000 log entries! > -Greg Er, whoops, as Ilya points out I've left out a step here. 100000 entries at 80 entries/sec is only 1250 seconds, or about 20 minutes. That's quite a different number and you should perhaps just ignore me. :)