Closed GoogleCodeExporter closed 9 years ago
Can you comment on the scp failures? Do you mean issue 143?
I monitor the response times and they're acceptable in the way that they are
comparable to when the server
doesn't have the extra load. During the extra load such as the generation of
images and doing the backup. The
load is still acceptable but I keep an eye on it.
I've adjusted the priority to medium as this will have to wait for some time.
You can already send me the info on
the move as you have it planned. Perhaps it won't be as big a deal as I am
afraid of;-)
Original comment by jurge...@gmail.com
on 20 Jan 2009 at 11:23
It's time consuming to split server and processing machine.
It's also unnecessary as the load is so limited. Only <100 per day requests by
Google analytics. I just added
you to the user list there, take a look.
That said, I'm ready to start migration as the 8 cores are just too tempting;-)
Who & what else is going to be on the machine besides this project?
I presume that the nmrrestrnts.bmrb.wisc.edu name will be transferred to it in
due time?
Original comment by jurge...@gmail.com
on 20 Jan 2009 at 1:04
Nothing else's supposed to be there.
Splitting is not about the load, it's about security and testing. So restraints
stay
internal until we copy it public server, and if something breaks we don't
publish the
broken stuff.
I'm not sure how it's supposed to work once we start processing deposited
restraint
files (as opposed to mr files sent to us by PDB), but if restraints get
processed and
published before main entry is released, I expect that might be a bad thing.
It's also not ideal that we have a public server (tang) on our intranet. We now
have
two webservers in a failover configuration, and they live in a completely
separate
subnet, with no user accounts, no access to nfs shares, and so on. I'm sure
they can
handle tomcat and < 100 requests/day -- and you get high-availability, software
updates
with no downtime and so on.
If we can mirror the database/servlet to public servers, nmrrestraints hostname
will
point to www.
Original comment by dmitri.m...@gmail.com
on 20 Jan 2009 at 10:36
Re: scp failures: scp encrypts everything, that takes CPU cycles. With load
hitting
100% on both CPUs, the cycles aren't available. Which may or may not be
contributing to
#143 problem.
Original comment by dmitri.m...@gmail.com
on 20 Jan 2009 at 11:51
Email me and Wim, details about the root access to this machine and I'll get
started on moving things over.
The reprocessing of files is going pretty slow which is a damn good reason to
start on this. I had love to see
the 8 cores hum;-)
Where would you suggest putting the db? It gets hit by processing and servlet
server. It's only a couple of Mb
of data in the whole db though. The actual data resides on the file system and
I am thinking of putting this on
the processing side. Then, can the server mount the partition(s) from the
processing side?
Write up your ideas please.
Original comment by jurge...@gmail.com
on 22 Jan 2009 at 6:58
No, I'd prefer a complete mirror of db and data, with no cross-mounts between
public
and private machines. We can mirror it with rsync the same way we mirror website
databases: csv dumps on private side, rsync daemon with a post-xfer script that
loads
db from csvs on public side.
As for root access, ssh to lionfish, from there ssh to grunt, then run "sudo su
-"
(works for both jurgen and wim accounts).
Original comment by dmitri.m...@gmail.com
on 22 Jan 2009 at 10:30
Cool, I am getting started to mirror tang's setup there. I think it's fine to
stay on FC9. Might not be a bad idea to
be a major version behind.
Can you setup the mirroring of the to be installed directories, db, and tomcat
servlet engine?
For the servlet engine you'll have to give me configuration details so that I
can adjust them in the Wattos.war.
Taking Eldon off the list as this gets too detailed.
Original comment by jurge...@gmail.com
on 23 Jan 2009 at 9:07
(I forgot how much I hate tomcat) Here's the deal: tomcat 5 does not work on
fedora 9.
Our choices are:
1. stay with fedora and use tomcat 6. I have no idea if it'll break anything in
your
servlet.
2. switch to CentOS 5.2: tomcat 5 works on it.
I installed tc6 on grunt, configs are in /etc/tomcat6, webapps in
/raid/www/webapps.
Drop your servlet in there and see if it runs... Mysql data dir is /raid/mysql.
I can set up the mirroring of data files and db, once you tell me the details:
what
files, where, db schema (I'm pretty sure I can dump it from mysql somehow) etc.
Original comment by dmitri.m...@gmail.com
on 24 Jan 2009 at 1:22
Also, grunt:/raid is mounted on tang as /grunt.
And tomcat-apache redirection seems to work: http://grunt/tomcat takes you to
the same
page as http://grunt:8080/
I'll probably need to open up the firewalls so you can see it from there -- or
you can
use ssh's dynamic port forwarding. Let me know.
Original comment by dmitri.m...@gmail.com
on 24 Jan 2009 at 1:35
So considering issue 165 I leave the initiative for this issue with you.
Original comment by jurge...@gmail.com
on 26 Jan 2009 at 2:10
Ok, let's keep the issue here now.
I have no reason to believe my code would fail on the newer tomcat but I
haven't tried it out yet. Since the code
is very small it would be easy to update for any backwards compatibility issues
you might find.
Original comment by jurge...@gmail.com
on 26 Jan 2009 at 8:15
Famous last words. I'm more worried about openjdk, though -- I see some weird
stuff
with validator UI. Anyway, if there are any compatibility issues you'll have to
update
the code sooner or later so it might as well be now.
Original comment by dmitri.m...@gmail.com
on 26 Jan 2009 at 8:34
OK, so what do I need to install to test the servlet and Java code?
Chris knows what's involved in processing, but that's not what I need. Is there
a
list of modules/projects with explanation of what they do and where to get them?
What's cinq, nmrrestraintsgrid, ccpn, and how it's different from wattos?
If you want to me to do this, you'll have to tell me what "this" is. Right now
I have
no idea where to start.
Original comment by dmitri.m...@gmail.com
on 27 Jan 2009 at 6:36
On second thought, apparently we don't care about any of this right now. So I'm
not
going to doing any of this anytime soon.
Original comment by dmitri.m...@gmail.com
on 27 Jan 2009 at 8:39
No problem.
Original comment by jurge...@gmail.com
on 28 Jan 2009 at 10:04
Original issue reported on code.google.com by
dmitri.m...@gmail.com
on 16 Jan 2009 at 8:17