<html> <head> <meta content="text/html; charset=windows-1252" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> <div class="moz-cite-prefix">On 19.05.2016 09:30, Stanislav Laznicka wrote: </div> <blockquote cite="mid:f57cd93a-0bc6-a338-4db2-84e88af528f6@redhat.com" type="cite">On 05/19/2016 08:52 AM, Ludwig Krispenz wrote: <blockquote type="cite"> On 05/19/2016 08:02 AM, Stanislav Laznicka wrote: <blockquote type="cite">On 05/18/2016 04:44 PM, Petr Vobornik wrote: <blockquote type="cite">On 05/18/2016 04:36 PM, Stanislav Laznicka wrote: <blockquote type="cite">There's no ticket for this patch but as there was a fix to 389-ds mentioned in <a class="moz-txt-link-freetext" href="https://fedorahosted.org/freeipa/ticket/5396">https://fedorahosted.org/freeipa/ticket/5396</a>, the TODO section in clean_dangling_ruvs could be removed. </blockquote> What about using 'replica-force-cleaning':'yes', every time? Is there a drawback which we would like to avoid? </blockquote> The DS website mentions two possible risks - possible loss of changes on deleted replica should these have not been reflected to some other replicas </blockquote> this is a theoretical concern that there might be changes from the replica to be removed which are not yet on all servers, but to me the problem that cleaning ruvs hangs because replicas cannot be reached is the worse scenario. <blockquote type="cite">- if some offline replica comes back online, it may re-pollute the RUVs back I'm not sure of the probability of the second scenario, in my rather simple environment the re-pollution did not happen. </blockquote> there have been fixes in 389-ds to prevent the repollution, so it should no longer happen. </blockquote> Thank you, Ludwig. It seems reasonable to have the option set to 'yes' all the time, then. <fieldset class="mimeAttachmentHeader"></fieldset> </blockquote> ACK Pushed to: master: 0492ab9c0a014735f09e82d5db1c4c1aa2bd6d81 ipa-4-3: d7985af911d54d3ea2637cd8200aeff5d66638d5 </body> </html>