GitLab

do virtnodes. Also change the query to exclude all nodes of class
'pcvm' from going into the ptop file.

The approach is also new, making use of the tmcd isalive status that
each node reports in. The fping stuff will get dropped as soon as I
make the changes to the rest of the testbed.

the testbed-ops mailing list for details.

This adds additional host routes to all ip addresses of a node
to be the shortest path one.

Still updating the old table though, until the mods are complete.

node_id,phys_nodeid to help keepalive stuff. Unfortuantely, this
one doesn't seem to be making much of a difference.

Also, add a pid,eid key to resrved to ehlp queries that want to
find nodes assigned to a certain project, for example.

to the virt_routes table, but it would delete them!

that although appeared to be right was actually causing errors to be
accumalated. This fixes the high error rates in the bandwidth and loss rate
numbers reported in OSDI. There are also code optimizations after profiling
it that reduces the RTSched overhead.

Another source of error was send/consume/request/reply which used to be
given a very rough estimate of the CPU speed (600 instead of 601.37 Mhz for
example). The latter comes from the boot up calibration in FreeBSD which
is supposed to be accurate up to 10 microsecs on a 486.  Using
600 instead of 601.37 causes an error of 0.22 % in the measurement. That is
about 1.3 ms for an RTT of 600 ms. This error is worse when send/consume
are used to calculate throughputs. The longer the period of measurement,
the worse it becomes.  I have committed changes in them as well. Defining
the macro CPU_SPEED_IN_KHZ will get you kernels that take CPU_SPEED
parameters in Khz instead of Mhz. So, you can specify 851940 instead of
850 for a pc850.

boss:/tftpboot/x86/{send,consume,request,reply}.cpuinkhz lets you specify
CPU_SPEED in Khz

updates that use it.

made the queries.

author of many of these queries. I mean, no one reads these commit
logs except Mike, right?

indicates who made the query (more or less, of course).

(results in a given experiment's hardware being returned as if it were free.)

that slow query log!

reserved table, since the vname is now stored in the reqp structure.
Speeds up those joins by quite a bit.

very interesting to look out and churns the logfile! Instead, just
print the number of routes/hosts returned.

always run.

Init $libdb::DBQUERY_MAXTRIES to 10 since the reload daemon should
always run.

but lets not take the chance.

If I didn't do anything in the last loop, wait 10 seconds instead of 1 before checking again. (It was hitting the db with 2 queries per second even when it wasn't doing anything.)

Order the query for the nodes by pid/gid, then save the info from the last
node. If the node I'm doing has the same pid/gid as the last one, reuse
the info without grabbing it from the database again and recalculating
everything. Now it will do one db query for each pid/gid that has an expt
swapped in, instead of once per node in use. That cuts the number of db
queries by a factor of (avg. # of nodes in use per proj), which is often
greater than 10.