Usenet admins....
Just curious folks. How big are the history files out there? The history.pag?
Ken.
In previous mail, Kenneth Whittaker sez...
Usenet admins....
Just curious folks. How big are the history files out there? The history.pag?
We average around 82-85 meg history.pag files with ~ 25 thousand groups.
John R. Dennison Senior Systems Administrator & Security Administrator WorldWide Access
Now that's awesome. How many inodes free on the filer? How does it perform?
John R. Dennison wrote:
In previous mail, Kenneth Whittaker sez...
Usenet admins....
Just curious folks. How big are the history files out there? The history.pag?
We average around 82-85 meg history.pag files with ~ 25 thousand groups. John R. Dennison Senior Systems Administrator & Security Administrator WorldWide Access
Kenneth Whittaker wrote:
Now that's awesome. How many inodes free on the filer? How does it perform?
Oops. My mistake. Not totally amazing. Is anyone running with a 1 GB history file? Anyone getting there? I know of one...
John R. Dennison wrote:
In previous mail, Kenneth Whittaker sez...
Usenet admins....
Just curious folks. How big are the history files out there? The history.pag?
We average around 82-85 meg history.pag files with ~ 25 thousand groups. John R. Dennison Senior Systems Administrator & Security Administrator WorldWide Access
-- Kenneth Whittaker Network Appliance Technical Support: 888-4-NETAPP
Chris Caputo wrote:
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
Oops. My mistake. Not totally amazing. Is anyone running with a 1 GB history file? Anyone getting there? I know of one...
Are people actually putting history files on the NetApp? Ours isn't.
Does it work?
Chris
I would not recommend that. I am guaging the sizes of the spools, directories, etc. The size of the history file seems to be a pretty good measure of the volume of news you are handling. Albiet, some people try to avoid rebuilding the history file by deleting the spool, but I don't know how much that buys us if the deletions themselves could be expensive.
I am assumeing the history files are not on the filer.
The size of the history file is not necessarily related to the amount of news held. This is because many history files contain message-ids for articles held in the past, long after they have expired off of the spool.
Some sites have 30 day history files even though they only hold articles for 10 days. As an example our history file currently contains:
Articles retained 3086663 Old entries retained 3342562
There are actually more lines devoted to message ids from the past than ones currently on the spool.
Chris
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
I would not recommend that. I am guaging the sizes of the spools, directories, etc. The size of the history file seems to be a pretty good measure of the volume of news you are handling. Albiet, some people try to avoid rebuilding the history file by deleting the spool, but I don't know how much that buys us if the deletions themselves could be expensive.
I am assumeing the history files are not on the filer.
-- Kenneth Whittaker Network Appliance Technical Support: 888-4-NETAPP
So, which would it imply if one finds a system with a 1 GB history file; 1) that the system does not rebuild the history file very often, or 2) that the system probably does a huge volume of news and cannot rebuild the history file often enough to keep the file resonably sized? Or, 3) other (please explain)?
Chris Caputo wrote:
The size of the history file is not necessarily related to the amount of news held. This is because many history files contain message-ids for articles held in the past, long after they have expired off of the spool.
Some sites have 30 day history files even though they only hold articles for 10 days. As an example our history file currently contains:
Articles retained 3086663 Old entries retained 3342562
There are actually more lines devoted to message ids from the past than ones currently on the spool.
Chris
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
I would not recommend that. I am guaging the sizes of the spools, directories, etc. The size of the history file seems to be a pretty good measure of the volume of news you are handling. Albiet, some people try to avoid rebuilding the history file by deleting the spool, but I don't know how much that buys us if the deletions themselves could be expensive.
I am assumeing the history files are not on the filer.
-- Kenneth Whittaker Network Appliance Technical Support: 888-4-NETAPP
If the admin is news saavy I would say 3) - the system is holding message ids for about a month and may or may not actually be holding news on the spool for that long. In INN this would be done by specifiying "/remember/:31" in the expire.ctl file.
If the admin is not news saavy I would say 1) or 2), but with a modification to 2) being that the software is not properly configured or they are not properly hardware equipped. It takes us about 26 minutes to process a .5 gig history file on a heavily loaded machine so it should take less than an hour for a full gig on a decently configured machine.
Chris
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
So, which would it imply if one finds a system with a 1 GB history file; 1) that the system does not rebuild the history file very often, or 2) that the system probably does a huge volume of news and cannot rebuild the history file often enough to keep the file resonably sized? Or, 3) other (please explain)?
Chris Caputo wrote:
The size of the history file is not necessarily related to the amount of news held. This is because many history files contain message-ids for articles held in the past, long after they have expired off of the spool.
Some sites have 30 day history files even though they only hold articles for 10 days. As an example our history file currently contains:
Articles retained 3086663 Old entries retained 3342562
There are actually more lines devoted to message ids from the past than ones currently on the spool.
Chris
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
I would not recommend that. I am guaging the sizes of the spools, directories, etc. The size of the history file seems to be a pretty good measure of the volume of news you are handling. Albiet, some people try to avoid rebuilding the history file by deleting the spool, but I don't know how much that buys us if the deletions themselves could be expensive.
I am assumeing the history files are not on the filer.
-- Kenneth Whittaker Network Appliance Technical Support: 888-4-NETAPP
-- Kenneth Whittaker Network Appliance Technical Support: 888-4-NETAPP
Chris Caputo wrote:
Are people actually putting history files on the NetApp?
Yes, for over a year now .
Does it work?
I would say so. What kind of history lookup times are you local-disk fanatics seeing?
For the poll:
-rw-rw-r-- 1 news news 732328706 Jun 23 15:31 history -rw-rw-r-- 1 news news 160599296 Jun 23 15:31 history.pag -rw-rw-r-- 1 news news 148 Jun 23 15:31 history.dir
Article lines processed 6479900 Articles retained 5968267
"News on NFS!? That's insane!"
Hmm... What kind of NFS OPS are you seeing? Typically I see (+/- a lot of ops) about 2000 by glancing at sysstat. I assume usenet filers have roughly the same mixure of OPS, but it could depend on configuaration as well I suppose.
Esa Ahola wrote:
Chris Caputo wrote:
Are people actually putting history files on the NetApp?
Yes, for over a year now .
Does it work?
I would say so. What kind of history lookup times are you local-disk fanatics seeing?
For the poll:
-rw-rw-r-- 1 news news 732328706 Jun 23 15:31 history -rw-rw-r-- 1 news news 160599296 Jun 23 15:31 history.pag -rw-rw-r-- 1 news news 148 Jun 23 15:31 history.dir
Article lines processed 6479900 Articles retained 5968267
"News on NFS!? That's insane!"
-- Esa Ahola esa@mindspring.net MindSpring Enterprises, Inc. 404.815.0770
On Thu, 19 Jun 1997, Kenneth Whittaker wrote:
Usenet admins.... Just curious folks. How big are the history files out there? The history.pag?
-rw-rw-r-- 1 news news 494156191 Jun 19 16:13 history -rw-rw-r-- 1 news news 70318400 Jun 19 16:12 history.pag
This is with a 15 day remember and an expire time anywhere from 3 to 14 days depending on the group.
Chris
Kenneth Whittaker penned:
Usenet admins.... Just curious folks. How big are the history files out there? The history.pag?
# ls -l history* -rw-rw-r-- 1 news news 275980551 Jun 20 09:10 history -rw-rw-r-- 1 news news 123 Jun 20 09:10 history.dir -rw-rw-r-- 1 news news 36733840 Jun 20 09:10 history.pag
# wc -l active 8947 active
# egrep -v '^#|^$' expire.ctl /remember/:14 *:A:4:10:never *.test:A:1:1:1 comp.*:A:4:14:16 news.*:A:4:14:never news.software.nntp:A:4:90:never news.announce.newgroups:A:4:75:never info.*:A:4:45:never gnu.*:A:4:30:never bit.*:A:4:18:18 sci.*:A:4:14:never ieee.*:A:4:30:never alt.*:A:7:7:7 alt.binaries*:A:1:1:1 alt.sex*:A:1:1:1 coin*:A:7:14:15 orion*:A:7:14:15 win*:A:7:14:15 control:U:1:1:1 junk:U:1:1:1
Regards,
David K. Drum david@more.net