From: Zdeněk Salvet Date: Tue, 15 Jul 2008 07:35:41 +0000 (+0000) Subject: IL backlogs, reference LB server tuning. X-Git-Tag: merge_316_6_dst~17 X-Git-Url: http://scientific.zcu.cz/git/?a=commitdiff_plain;h=b968c18061bf9c1ff08fc4919184897632fdee2c;p=jra1mw.git IL backlogs, reference LB server tuning. --- diff --git a/org.glite.lb.doc/src/LBAG-Installation.tex b/org.glite.lb.doc/src/LBAG-Installation.tex index 15a1a6f..b9be338 100644 --- a/org.glite.lb.doc/src/LBAG-Installation.tex +++ b/org.glite.lb.doc/src/LBAG-Installation.tex @@ -13,6 +13,7 @@ popisky rovnou doplnit do RPM \subsection{\LB server} \subsubsection{Hardware requirements} +\label{inst:hw_req} Hardware requirements depend on performance and storage time requirements. Disk space used by LB server consists of database space and working space @@ -117,6 +118,7 @@ The default startup script checks for existence of \subsubsection{Exploiting parallelism} \subsubsection{Tuning database engine} +\label{inst:db_tuning} In order to achieve high performance with LB server underlaying MySQL database server has to be configured reasonably well too. diff --git a/org.glite.lb.doc/src/LBAG-Running.tex b/org.glite.lb.doc/src/LBAG-Running.tex index 51ff12d..e3ee44e 100644 --- a/org.glite.lb.doc/src/LBAG-Running.tex +++ b/org.glite.lb.doc/src/LBAG-Running.tex @@ -343,14 +343,18 @@ for glite-lb-interlogd to pick them up again \paragraph{Slow delivery.} Either glite-lb-interlogd or the target \LB server(s) may not keep pace -with the incoming stream of events. -Unless the situation is permanent, no specific action is required---the +with the incoming stream of events. +Amount of such backlog can be quickly assessed by looking at timestamps of the oldest +event files (with \verb'ls -lt $GLITE_LOCATION_VAR/log/', for example). +Fully processed event files are deleted in approx. one minute intervals, +so files last modified one minute ago or newer do not constitute backlog at all. +Unless the backlog situation is permanent, no specific action is required---the event backlog decreases once the source is drained. Otherwise hardware bottlenecks (CPU, disk, network) have to be identified -(with standard OS monitoring) and removed. - -\TODO{salvet -- nemame jeste nejake rychle tipy?} - +(with standard OS monitoring) and removed, see sections~\ref{inst:hw_req} and +\ref{inst:db_tuning} for \LB server tuning tips. In order to maximize +performance of \verb'glite-lb-interlogd', it is recommended to distribute +jobs to multiple \LB servers that can be shared by multiple WMS setups in turn. \subsubsection{Notification delivery} \begin{sloppypar}