From d415d4d32f2e8e61de21abfdfce02e1b1ea1e1d3 Mon Sep 17 00:00:00 2001 From: Volker Lendecke Date: Tue, 3 Nov 2009 05:41:02 +0100 Subject: s3: Add parameter "ctdb timeout" When something in the cluster blocks, it can happen that we wait indefinitely long for ctdb, just adding to the blocking condition. In theory, nothing should block, but as someone said "In practice the difference between theory and practice is larger than in theory". This adds a timeout parameter in seconds, after which we stop waiting for ctdb and panic. --- docs-xml/smbdotconf/misc/ctdbtimeout.xml | 37 ++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) create mode 100644 docs-xml/smbdotconf/misc/ctdbtimeout.xml (limited to 'docs-xml/smbdotconf/misc') diff --git a/docs-xml/smbdotconf/misc/ctdbtimeout.xml b/docs-xml/smbdotconf/misc/ctdbtimeout.xml new file mode 100644 index 00000000000..97d50390000 --- /dev/null +++ b/docs-xml/smbdotconf/misc/ctdbtimeout.xml @@ -0,0 +1,37 @@ + + + This parameter specifies a timeout in seconds for the + connection between Samba and ctdb. It is only valid if you + have compiled Samba with clustering and if you have + set clustering=yes. + + When something in the cluster blocks, it can happen that + we wait indefinitely long for ctdb, just adding to the + blocking condition. In a well-running cluster this should + never happen, but there are too many components in a cluster + that might have hickups. Choosing the right balance for this + value is very tricky, because on a busy cluster long service + times to transfer something across the cluster might be + valid. Setting it too short will degrade the service your + cluster presents, setting it too long might make the cluster + itself not recover from something severely broken for too + long. + + + Be aware that if you set this parameter, this needs to be in + the file smb.conf, it is not really helpful to put this into + a registry configuration (typical on a cluster), because to + access the registry contact to ctdb is requred. + + Setting ctdb timeout to n makes + any process waiting longer than n seconds for a reply by the + cluster panic. Setting it to 0 (the default) makes Samba + block forever, which is the highly recommended default. + + +0 + -- cgit v1.2.1