Server Status Page - What do those columns mean?

[anandhanju]

Can any of the site mods help me out with the abbreviations on the server stats page and what they indicate? I understand that the things needed for interpreting useful info (such as Status and Netload) are provided on the page but I sure am curious what those other whatchamacallits are.

The *ed ones are those that are typically checked by a Folder if a result cannot be upload/ new WU cannot be received.

The texts in italicized red are guesses.

Note: The observations noted below may not be accurate as long as you see this warning!

SERVER IP* - IP address of FAH server.
CLIENT -The type of projects this server is currently handling.

Code: Select all

"-" - server is an Assignment Server or a Collection server i.e., it does not hand out work.
"classic" - Classic work units for the single-processor CPU clients.
"GPU" - GPU2 work units for the ATI/ nVidia client.
"PS3" - PS3 work units.
"SMP" - SMP work units for the SMP (MPICH/ Deino) clients.

NAME - Name of FAH server
WHO - FAH researcher who "owns" the server
STATUS* - Current state of the server.

Code: Select all

"Full" - server is in full operation, everying is ok.
"Accept" - server is only accepting WUs, not assigning.
"Fail" - server can't be reached by the stats system.
"DOWN" - server is down.
"STANDBY" - server is not serving FAH WUs.
Special values: "AS x" - Assignment server; "CS x" - Collection server; "internal use" - For internal testing purposes

CONNECT* - Current result of our serverstatus code to try to connect to the F@H server.

Code: Select all

"Accepting" - server is alive and well and able to accept client communication (TCP connections).
"Reject" - server should be up and accepting connections, but isn't.
"Not Accept" - server is at standby and we don't expect it to be accepting connections (and it isn't).
"OFFLINE" - server is offline for planned maintenance.

CPU LOAD* - Load on the server (# of processes). Anything more than 3-4 is bad
NET LOAD* - # of current connections the server is handling. Too many connections means that the server is heavily loaded. How many are "too many" depends on the server, but most of the servers can now handle a couple hundred connections without a problem.
FIN WAIT - ?
DL - Days left on at least one of the partitions
GB TOT - Total disk space
GB AV - Available disk space
DIFF TIME - ?
WU L - ?
WUs AVAIL - # of WUs that are available for the server to give out
WUs to go - ?
WUs WAIT - Shows how many WUs have been assigned by the data server. This includes WUs which are beyond the deadline, so it may give an exaggerated count.
% Ass - % of last 500 assignments on port 8080
% Ass 80 - % of last 500 assignments on port 80
% Ass G - % of last 500 GPU assignments (on port 8080)
% Ass PS - % of last 500 PS3 assignments since the last update (on port 8080)
WUs RCV - rolling total of # of WUs received by the server since the last update, per 2 hours.
WUs E - Number of WUs that have "expired" and will be reissued
T - ?
st - ?
S - ?
CS - See VijayPande's post below.
CSListed - See VijayPande's post below.
NMJ - No more jobs? Woohoo!
80 - True or false indication about whether the server is redirecting port 80 connections to server port 8080 (i.e., is port 80 working) - bruce
OperatingSystem - Operating Systems that this server can assign work to.

Code: Select all

W = Windows
L = Linux
X = Mac OS X

WEIGHT - "Weight" of the server (how frequently the AS will assign clients to that particular server). Higher numbers means more frequently. Zero means never.
min ver - Minimum client version to get work from that server - 7im
Min_packet - Minimum Packet size of WUs (for deciding small/medium/big [BigWU] assignments)
Max_packet - Maximum Packet size of WUs (for deciding small/medium/big [BigWU] assignments)
memory - Minimum memory required, as reported by the fah client, to be eligible to get work from that server. - 7im
smp cores - # of cores for SMP WU assignment
gp type - GPU Type? Possible values - 1, 2
PROGRAM - Type of projects currently handled by the server.

Code: Select all

F - FAH projects for everyone.
A - Advmethods projects, require the -advmethods flag to be enabled.
B - Beta projects, for testing purposes only. Should move to F or A shortly.
I - Internal projects, for testing purposes only. Should move to F or A, but might be only experimental WU that will never be released.
G - Used to be GAH projects, no longer used.

AssignedPort - Ports being used for assigning/receiving work
ver - Probably a number that increases as changes are made to the server code. - bruce
C - ?
T - ?
G - ?
RMEM - All 0s...
WHO - (Duplicate column) FAH researcher who "owns" the server
PBL - ?
NAME - (Duplicate column) Name of FAH server

InCS and TC column is gone. Added CS, CSListed, WHO, and Name columns. ~2008/05. -7im

[bruce]

There's no information available except what is written at the bottom of the page.

You've done a pretty good job of guessing (at least some of your guesses correspond to my guesses, so I choose to call that "good" whether we're both right or both wrong).

80 - True or false indication about whether the server is redirecting port 80 connections to server port 8080 (i.e.- is port 80 working)

ver - (Probably a number that increases as changes are made to the server code.)

[7im]

The one I know from asking questions of the Pande Group in the past is the Memory column. That is the minimum memory required, as reported by the fah client, to be eligible to get work from that server. If that column says 300 MB, and your old P3 only has 256 MB, you won't get work from that server.

Min Ver is the minimum client version to get work from that server.

Assigned Port is easy to guess. The client only works over ports 80 and 8080, so that matches the settings in those columns. It's also spelled out in the legend below that Server Status page. You might want to double check it.

[anandhanju]

The first post has been updated to remove the recently deleted columns in the server stats page.

@FAH team, Thanks for adding the column headers are regular intervals. How long will historic logs be displayed for a server? IIRC, it was 2 days but today I see records since the past 5 days (E.g., http://fah-web.stanford.edu/logs/171.65.103.93.log.html)

[toTOW]

PROGRAM - ?

Code: Select all

F - FAH projects for everyone.
A - Advmethods projects, require the -advmethods flag to be enabled.
B - Beta projects, for testing purposes only. Should move to F or A shortly.
I - Internal projects, for testing purposes only. Should move to F or A, but might be only experimental WU that will never be released.
G - Used to be GAH projects, no longer used.

My 2 cents

[VijayPande]

We've been sprucing up serversstat by adding new columns, deleting old ones which we didn't use, and speeding up the update code (by about 2x). The main addition has been better collection server (CS) info. In the past, FAH had just one CS, so the question was "is this server set up correctly for *the* CS", but now with multiple-CS's, there are 2 questions: "Does the CS know about this server and is this server set up for the right CS." The serverstat has been updated to give that info. You'll see there are a few issues and my team is working on them now that they're more obvious (ETA: Tuesday). We're also working to setup a new CS for SMP WU's only.

[EvilAlchemist]

Yeah, I have always wondered what "WUs to go " means as well.

[m58@itpro]

Check this correction ?

% Ass - % of last 500 assignments on port 8080 // Correction: This is just total percent assigned to server from main AS, forget the Port 8080 ?
% Ass 80 - % of last 500 assignments on port 80 // Correction: The "80" means 8086-architecture (classic or SMP) ?
% Ass G - % of last 500 GPU assignments (on port 8080) // Correct, but forget the port 8080 ?
% Ass PS - % of last 500 PS3 assignments since the last update (on port 8080) // Correct, but forget the port 8080 ?

The assignment data:
Assignment total, the percent assigned to this server
Assignment from 8086 assignment server (why should they specify port 80 here ?)
Assignment from GPU assignment server
Assignment from PS3 assignment server

[7im]

Where are you getting this corrected info from? 8086 has nothing to do with assignments. 80 means port 80. 8080 means port 8080.

[bruce]

Until recently there were four Assignment Servers:
AS
AS 80
AS (GPU)
AS (PS3)
The percentages you're asking about represent which work servers were recently assigned by each of those Assignment Servers.

The primary AS listens only on port 8080. The secondary AS listens on both port 80 and 8080 but because the client tries the primary AS first, AS2 (or AS80) generally only handles requests through port 80. Both servers handle classic CPU and SMP clients. The GPU and the PS3 clients contact their own Assignment Servers.

NOTE: Because of how the data is collected, the values are approximate.

It would not surprise me to find that this information has become obsolete at any time. The Pande Group does not announce changes to the assignment process.

AS PS3B been added recently and I'm not sure if there is a way to tell what it's doing.

[anandhanju]

Serverstat for server 171.64.65.60 reports the same number of WUs for WUs AVAIL, WUs to go, WUs WAIT. While AVAIL/ go could be the same, it is highly unlikely the WAIT is always the same as the other two. It is a minor thing, but it seems to be occurs for other servers as well (129.74.85.48, 128.143.48.226, 171.67.108.20, 171.67.108.21, 171.64.65.71).

These appear to be newer servers and the common thing tying these servers (from what I can see on serverstat) is that they do not report a version under the Ver column. Maybe someone can take a look when they have some time to kill (I know, VJ is here at midnight posting that servers are back online, jcoffland is juggling multiple new cores and Beberg's Tuesday is fast approaching -- these FAH researchers have all the time in the world )

[gwildperson]

These appear to be newer servers and the common thing tying these servers (from what I can see on serverstat) is that they do not report a version under the Ver column. Maybe someone can take a look when they have some time to kill (I know, VJ is here at midnight posting that servers are back online, jcoffland is juggling multiple new cores and Beberg's Tuesday is fast approaching -- these FAH researchers have all the time in the world )

. . . and you can be sure that Kasson is busy with new developments, too, so he's got all the time in the world, just like the others.

[susato]

Today, servers vspg4, vspg4v and vspg4v2 have "am" in the DL (days left) column. That's a new one.