mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PrimeNet (https://www.mersenneforum.org/forumdisplay.php?f=11)
-   -   Primenet Server - Official Maintenance Thread (https://www.mersenneforum.org/showthread.php?t=21648)

Madpoo 2016-10-11 21:25

Primenet Server - Official Maintenance Thread
 
I'm starting a new thread to announce server maintenance tasks.

To kick things off, our collocation host has given us a 3-day window during which they will be physically moving our server from one place to another on the same floor.

That maintenance is scheduled to occur during any of these 3 windows:
[CODE]Tue 2016-10-11 between 23:00 UTC and 07:00 UTC the following day
Wed 2016-10-12 between 23:00 UTC and 07:00 UTC the following day
Thu 2016-10-13 between 23:00 UTC and 07:00 UTC the following day
[/CODE]

Unfortunately that's as specific as it gets, so I'm just expecting it on any of those days. The actual length of outage is estimated to be 60-100 minutes once they start.

Once the maintenance is complete I'll follow up so we know it's done.

Mark Rose 2016-10-11 22:35

Oooo.... server migration :D

Madpoo 2016-10-12 01:01

[QUOTE=Mark Rose;444816]Oooo.... server migration :D[/QUOTE]

Yeah, a real hoot and a holler. :smile:

Just got forwarded an email from George that apparently *tonight* is the night. I'm figuring on whether I should do a graceful shutdown ahead of time or have them hit the power button and let it save the virtual machine's state on it's own. Decisions decisions...

I'll be kicking off a fresh backup of the data and hope it syncs up to the cloud beforehand, at any rate. We'll see. It should all be good though.

snme2pm1 2016-10-12 06:39

[QUOTE=Madpoo;444823]*tonight* is the night. I'm figuring on whether I should do a graceful shutdown ahead of time or have them hit the power button and let it save the virtual machine's state on it's own.
[/QUOTE]

Hell of a way to test fail safe provisions!
My notion of "tonight" is somewhat different to yours I would expect.
mersenne.org seems quiet now...
Ed: Assuming decent arithmetic, UTC 07:02 apparently resumed sometime earlier. (I wasn't watching closely).

Madpoo 2016-10-12 14:20

[QUOTE=snme2pm1;444837]Hell of a way to test fail safe provisions!
My notion of "tonight" is somewhat different to yours I would expect.
mersenne.org seems quiet now...
Ed: Assuming decent arithmetic, UTC 07:02 apparently resumed sometime earlier. (I wasn't watching closely).[/QUOTE]

I got tired of waiting and went to bed... looks like it happened between 11:30 PM and midnight'ish in my timezone (PDT). (EDIT: 11:23 PM to 11:57 PM according to my Pingdom monitor)

It worked out pretty well. The virtual machine saved state and came back up okay. I'm now doing a reboot of the server since I don't always trust the saved-state as far as OS stability... when it works it works usually but I've had experiences with hung services or other oddities.

EDIT: And we're back up. Over all a pretty decent process. It helped that it was apparently just from one spot on the same floor to another. I've done entire cabinet/cage relocations and it's not fun.

Madpoo 2016-12-22 22:39

Server reboot coming up
 
The Primenet server is going to get a reboot shortly. The online backup service had stalled out and it just bugged me about it, so to be sure I'm going to reboot the server. I've seen that happen before when the service (Carbonite in this case) updates the client software and although it doesn't say so, it's best to reboot it since files were in use, blah blah. Windows stuff.

It should be a quick thing... I'm also updating the AV and a few other odds and ends as long as I'm at it.

Time is now 22:38 UTC and I'm planning to hit the button around 22:40 UTC. Should only be about 5-10 minutes of downtime.

EDIT: Back online at about 22:50 UTC.

Madpoo 2016-12-23 16:08

Minor SQL update occurring now
 
I'm doing a quick SQL update on Primenet. It failed to apply automatically yesterday during the maintenance... I already knew I can never trust Windows Update to do SQL updates... it always seems to fail. Don't know why I thought it'd work yesterday either. :smile:

EDIT: Back online. Easy breezy.

Madpoo 2017-01-18 21:39

Quick server reboot coming up - prepping for SSL
 
I added a certificate to Primenet and now I need to adjust the horrible default cipher suite that Microsoft uses. You'd think they'd disable things like sslv3 in some security hotfix, but I guess they're afraid that some website out there would freak out if that happened.

I'm shooting for an "A" score on ssl labs... we'll see.

Outage will begin now and should only last a few minutes.

Madpoo 2017-01-18 21:44

[QUOTE=Madpoo;451184]...
I'm shooting for an "A" score on ssl labs... we'll see.
...[/QUOTE]

:smile:

[URL="https://www.ssllabs.com/ssltest/analyze.html?d=www.mersenne.org&latest"]www.mersenne.org SSL test[/URL]

Dubslow 2017-01-19 07:27

[QUOTE=Madpoo;451185]:smile:

[URL="https://www.ssllabs.com/ssltest/analyze.html?d=www.mersenne.org&latest"]www.mersenne.org SSL test[/URL][/QUOTE]

Is it possible to force all incoming connections to https by default?

S485122 2017-01-19 20:27

[QUOTE=Dubslow;451209]Is it possible to force all incoming connections to https by default?[/QUOTE]Yes it is. But for the moment HTTPS is in test phase (I do use it and it works perfectly for me.)

Mark Rose 2017-01-19 20:33

Who knows what the effect may be on the various spiders that connect to the HTTP site.

Dubslow 2017-01-20 06:48

[QUOTE=Mark Rose;451233]Who knows what the effect may be on the various spiders that connect to the HTTP site.[/QUOTE]

Okay, maybe "force" was wrong -- is there a way to inquire a new connection if it is capable of switching, and if so, to do so?

Mark Rose 2017-01-20 21:40

[QUOTE=Dubslow;451250]Okay, maybe "force" was wrong -- is there a way to inquire a new connection if it is capable of switching, and if so, to do so?[/QUOTE]

The best way may be with a script at the top of the document. The spiders probably aren't executing JS.

Madpoo 2017-01-21 03:10

[QUOTE=Mark Rose;451233]Who knows what the effect may be on the various spiders that connect to the HTTP site.[/QUOTE]

That's the biggest concern. Regular web browsers won't be affected, except for the warm fuzzies they get from seeing that it's a secure connection.

I locked down the cipher suite to make sure that when we go secure, we're not doing it half-heartedly. That means obvious things like disabling SSLv3 and basically going with TLS and using only the ciphers with forward secrecy. The SSL Labs test gives it an A and the downside is if you're from the stone age and visiting with Windows XP and IE 6, you're out of luck. All other modern-ish browsers (and I think even the old funky Firefox 3.x that a certain someone prefers) should be okay.

But yeah... the people who use scripts to crawl result pages or collect the reports... if I start redirecting http -> https (which is easy to do, by the way), depending on how they did their script it may ignore a 301/302 redirect, or it may not be setup to work with SSL.

I know curl by itself would need a list of trusted CAs (or use the -k option to ignore cert issues, which isn't the best idea). I really don't know what people are using to crawl the site with so we'll probably just have to give them time to test it and then make the switch.

I started a new thread devoted to discussing the SSL switch so I'll be monitoring things there.

Prime95 2017-01-21 04:53

Most importantly, be sure the prime95 client which communicates using http is not affected by any changes.

Madpoo 2017-01-24 03:54

[QUOTE=Prime95;451309]Most importantly, be sure the prime95 client which communicates using http is not affected by any changes.[/QUOTE]

Good point. The client communicates with v5.mersenne.org so it's fortunately separate from the website.

With that said, it might be a good future project to get new clients to talk over SSL as well, although it's probably not critical. Passwords aren't passed along using the API (as far as I'm aware).

At any rate, SSL on the website should leave the Prime95 clients alone. The biggest concerns I had were with GPU72 and Misfit since (not sure on the details) they can proxy client activity and then talk to the manual assignment/result pages, or something like that. I'm fuzzy on just how those worked, thus my concern.

It's not terribly difficult to exclude the manual assign/result pages from being redirected, if there's still a concern about that.

retina 2017-01-24 05:24

[QUOTE=Madpoo;451308]All other modern-ish browsers (and I think even the old funky Firefox 3.x that a certain someone prefers) should be okay.[/QUOTE]I never have any trouble with other sites, so unless you have done something particularly restrictive then it should be fine.

chalsall 2017-01-24 15:48

[QUOTE=Madpoo;451471]Passwords aren't passed along using the API (as far as I'm aware).[/QUOTE]

That is correct.

The worst someone could do who was "sniffing the plain-text traffic" would be to replay assignment requests. If they reverse engineered the client's "secret sauce" security code they would also be able to unassign assignments.

But really, that's an exceptionally unlikely scenario. Someone would have to really hate someone else to get "in the middle", and also invest a great deal of time and effort. And even then they wouldn't be able to log into Primenet's web interface as their target's user (they would have the Username, but not the Password).

Madpoo 2017-01-26 15:51

[QUOTE=retina;451476]I never have any trouble with other sites, so unless you have done something particularly restrictive then it should be fine.[/QUOTE]

Give it a try then: [url]https://www.mersenne.org[/url]

If "even Retina" can access it, then I think we can consider it effectively 100% browser compatible. :smile:

henryzz 2017-01-26 15:57

[QUOTE=Madpoo;451603]Give it a try then: [URL]https://www.mersenne.org[/URL]

If "even Retina" can access it, then I think we can consider it effectively 100% browser compatible. :smile:[/QUOTE]
IE6? Was very popular once. I suppose you should be used to webpages not working on IE6 by now.

retina 2017-01-27 01:39

[QUOTE=Madpoo;451603]Give it a try then: [url]https://www.mersenne.org[/url]

If "even Retina" can access it, then I think we can consider it effectively 100% browser compatible. :smile:[/QUOTE]Okay, it looks as though you have effectively achieved 100% browser compatibility. :tu:

Madpoo 2017-05-12 23:48

Server reboot - happening now
 
I'm about to reboot the Primenet server to make sure updates are installed. The ransomware outbreak going on today had me spooked and even though the server is firewalled these things make me nervous when the server hasn't been updated in a couple months.

As usual I'll say it'll only take 10 minutes but it will really take 20. So I'm saying it'll take 20 minutes (which means it might take 30?) :smile:

EDIT: And, back online after just 15 minutes or so. All done.

Madpoo 2017-07-22 06:09

Primenet reboot coming up
 
I need to do some offline maintenance of Primenet... I did a little SQL shuffling of files and now I need to shut it down and make some changes on the virtualization layer.

I'm hoping it'll just be a quick process, maybe 10 minutes downtime max, but I can't really say how long one of the jobs will take.

Sorry for the interruption, but we'll have it back up and going before you know it.

EDIT: It finished pretty quick... back up and running.

Madpoo 2017-09-09 02:59

Server maintenance - happening now
 
The Primenet server is going offline right now for about 30-60 minutes.

The plan is to make a clone of the system to create a testbed for work on PRP handling code (and we can also, in the future, use it for other system upgrades where we don't want to break the live system accidentally).

I'm not sure how long it'll take to make copies of the virtual drives so I can't be more specific... hopefully faster than an hour.

Madpoo 2017-09-09 04:10

[QUOTE=Madpoo;467446]The Primenet server is going offline right now for about 30-60 minutes.

The plan is to make a clone of the system to create a testbed for work on PRP handling code (and we can also, in the future, use it for other system upgrades where we don't want to break the live system accidentally).

I'm not sure how long it'll take to make copies of the virtual drives so I can't be more specific... hopefully faster than an hour.[/QUOTE]

Back online... enjoy. :smile:

Madpoo 2017-10-14 18:24

Maintenance starting soon
 
I'll be performing some server maintenance soon... a few minor updates but the big one will be updating SQL which will take longer.

I'm planning a downtime of 30-60 minutes but hopefully it'll be much less. Weekends are the time for doing such things. :smile:

Starting at approx 18:30 UTC (~ 5 minutes from now)

Madpoo 2017-10-14 19:52

[QUOTE=Madpoo;469845]I'll be performing some server maintenance soon... a few minor updates but the big one will be updating SQL which will take longer.

I'm planning a downtime of 30-60 minutes but hopefully it'll be much less. Weekends are the time for doing such things. :smile:

Starting at approx 18:30 UTC (~ 5 minutes from now)[/QUOTE]

All done. I took a bit over an hour, but with a couple periods of uptime in between. Good to go.

Madpoo 2017-10-15 05:34

Quick offline maintenance
 
The SQL upgrade revealed some horribly fragmented indexes... took the server offline real quick to rebuild a few that were slowing things down. Should only be a few minutes (knock on wood).

ric 2017-10-15 08:11

[QUOTE=Madpoo;469867]The SQL upgrade revealed some horribly fragmented indexes... [/QUOTE]

This would explain why I'm experiencing [CODE]Fatal error: Maximum execution time of 300 seconds exceeded in C:\inetpub\www\report_exponent\default.php on line 275[/CODE] and/or [CODE]500 - Internal server error.
There is a problem with the resource you are looking for, and it cannot be displayed.[/CODE] while searching for a single exponent. Hope it's transitory...

GP2 2017-10-15 09:22

[QUOTE=ric;469871]while searching for a single exponent. Hope it's transitory...[/QUOTE]

Yes, checking the exponent status on a single exponent can take a minute or more. Even if "Show full details" is turned off, it can take up to ten seconds.

Seems like some database indexes are not complete.

Madpoo 2017-10-15 16:02

[QUOTE=GP2;469872]Yes, checking the exponent status on a single exponent can take a minute or more. Even if "Show full details" is turned off, it can take up to ten seconds.

Seems like some database indexes are not complete.[/QUOTE]

Yeah, I'm running into some horrible query plans. The log file also filled up overnight which was... unexpected. Probably my fault because of all the stuff I'd done and wasn't paying attention to it. Woke up to "out of space"... bleagh.

Fixed, and now I can troubleshoot the poor queries. :smile: To be honest, it actually seemed a little better because I nuked the cached query plans while I was at it. Maybe it got its head out of its butt. LOL

GP2 2017-10-16 02:12

The [URL="https://mersenne.org/assignments/"]Active Assignments page[/URL] is empty (no assignments listed).

Exponent Status is still mostly much slower than before.

kruoli 2017-10-16 09:50

[QUOTE=GP2;469913]The [URL="https://mersenne.org/assignments/"]Active Assignments page[/URL] is empty (no assignments listed).[/QUOTE]

On my end, that's because the maximum exponent is now set to 2 by default. If I change it to 10000 for example, it displays fine ([url]https://www.mersenne.org/assignments/?exp_lo=2&exp_hi=10000[/url]).

Madpoo 2017-10-17 01:58

[QUOTE=kruoli;469931]On my end, that's because the maximum exponent is now set to 2 by default. If I change it to 10000 for example, it displays fine ([url]https://www.mersenne.org/assignments/?exp_lo=2&exp_hi=10000[/url]).[/QUOTE]

I'm pretty sure it's been set at 2 and 2 for low/high for a long time.

Sorry for the way the exponent report page has been performing the past few days.

I was just working on it (and accidentally left in a debug code that showed me the raw SQL statement...whoops).

I think I've narrowed down the problem to a few possibilities and have hopefully worked around it. The upgrade to SQL apparently took things down a dark path... where some queries used to work okay, they were actually kind of funky and I think SQL just had a hard time working out the past execution plan for them.

Rewriting some of the bits on that report seems to have improved things. Hard to say for sure yet because sometimes it'd run nice and fast, sometimes it was taking 60+ seconds for a single exponent, which really bugged me. Nothing worse than trying to troubleshoot an inconsistent problem.

I have one more technique I can try if needed, and in my testing it cut down the query time even more, so I may do it anyway (using CTEs to limit the exponent set beforehand), but I wanted to see how the stuff I did already is performing.

I still want to do a *total* rewrite of that report because it's just kind of top-heavy anyway in the way it collects all the data. That's why the XML version of that page runs so dang fast, because it's using my own re-written query (which I need to update to include PRP stuff, now that I think about it).

Madpoo 2017-10-17 04:12

Credit goes to George for updating that query to basically tell SQL to "just do it in the order it's written" since apparently it was trying to overthink it and shooting itself in the foot.

I came up with an alternate strategy that seemed to accomplish the same thing, but it was relying on SQL to "do the right thing" which you really can't rely on apparently. :smile:

Madpoo 2017-11-25 22:44

Maintenance happening now
 
I've taken the Primenet server offline for a quick round of maintenance. I decided it was time to slightly grow one of the virtual disks for the SQL log files. It's not uncommon these days for the log to grow kind of large between full backups and I just wanted to give ourselves more buffer to make sure it's not an issue.

Then just routine updates while I'm at it... back up shortly.

EDIT: Back online

henryzz 2017-11-29 11:36

[QUOTE=Madpoo;472431]I've taken the Primenet server offline for a quick round of maintenance. I decided it was time to slightly grow one of the virtual disks for the SQL log files. It's not uncommon these days for the log to grow kind of large between full backups and I just wanted to give ourselves more buffer to make sure it's not an issue.

Then just routine updates while I'm at it... back up shortly.

EDIT: Back online[/QUOTE]

You don't have autogrowing virtual disks?

Madpoo 2018-02-08 19:51

Server maintenance - 2018-02-18
 
The Primenet server is about to get a reboot following some OS/db patches. Should be a short maintenance period.

EDIT: done

Madpoo 2018-04-01 18:31

upcoming system maintenance - 2018-04-01
 
Not an April fools joke... :smile:

I'll be doing a handful of updates to the server with a couple reboots.

EDIT: done

Madpoo 2018-06-14 03:04

Server maintenance - 2018-06-14 @ 3:15 AM UTC
 
Taking the server offline for a brief maintenance period... will only be about 20-25 minutes.

-- 3:32 AM UTC - all done

Madpoo 2018-08-05 05:51

Maintenance - 2018-08-05
 
Routine maintenance is happening now... sorry I didn't announce it ahead of time. Just some SQL and other Windows updates.

Should be another 20 minutes or so from now.

EDIT: All done

Madpoo 2018-10-20 17:00

Regular maintenance - 2018-10-20 @ 17:00
 
The server will be down temporarily for routine maintenance - as usual this may take 20-30 minutes.

UPDATE: all done - looking good.

Madpoo 2018-12-08 07:59

I'm about to kick off some regular maintenance. I'd normally let it ride but I forgot to install some SQL updates last time and I wanted to make sure I got those in.

I'm estimating 35 minutes to complete. Starting at 8 AM UTC, should hopefully be done by 8:35'ish UTC

EDIT: all done, 5 minutes early :smile:

Madpoo 2018-12-28 05:11

Doing a reboot of the server soon - updating the backup software, so that's important. :smile:

It's 2018-12-28 @ 5:11 AM with an expected downtime ~ 15 minutes.

EDIT: Back online. Okay, it was more like 18 minutes.

Madpoo 2019-02-04 03:09

Maintenance scheduled - 2019-02-04 @ 03:15 (UTC)
 
I'm going to go ahead and do some updates on the Primenet server:
2019-02-04 @ 03:15 (UTC)

I noticed a funny period of SQL queries taking long and even though it's stabilized (and only lasted a few minutes about 30 hours ago) I wanted to go ahead and update things.

I'm only doing SQL updates so the downtime should be minimal, ~ 5 minutes.

EDIT: yup, it was about 5 minutes, all done.

Madpoo 2019-03-24 05:48

Maintenance about to start
 
I'm about to do some updating on the server.

It'll be down probably 30 minutes, and I'm starting at 05:50 UTC on 2019-03-24

EDIT: All done

Madpoo 2019-04-20 16:44

Maintenance - 2019-04-20 @ 16:40 UTC
 
I'm taking the server offline for a few minutes for some SQL related maintenance.

I could have done it offline but I'm altering some tables that would make it slow and non-responsive anyway which would only be more frustrating for people. :smile:

I expect it to finish in a few more minutes.

EDIT: Back in action. Took longer to update that pesky table than I had hoped... probably would have been faster to drop the indices, alter it, then recreate indices. Oh well. All done.

Madpoo 2019-04-21 06:58

More SQL maintenance - 2019-04-21 @ 06:55 UTC
 
There is more SQL maintenance happening right now. At least this time I waited until later in the evening in the US.

Working on some tables, dropping and adding columns. It affects web visits so I disabled the web service while the work is happening.

ETA about 25 minutes

EDIT: Back online - didn't take as long as I thought.

Madpoo 2019-05-15 18:04

May 2019 maintenance
 
I'm going to be taking the server offline for a brief period to do some routine maintenance.

I expect to start @ 18:10 UTC (2019-05-15) and it should be ~ 20 minutes (it's now 18:05 UTC, so yeah... 5 minutes from now)

UPDATE: It finished a little ahead of time (by a few minutes). Good to go.

kriesel 2019-05-15 22:24

Is this what happens during server maintenance?
 
1 Attachment(s)
In response to <link removed by LaurV>

I got a null page (no assignments, no error message)
Or maybe it relates to my slow but unreliable DSL.

Madpoo 2019-05-18 16:41

[QUOTE=kriesel;516868]In response to <url removed>

I got a null page (no assignments, no error message)
Or maybe it relates to my slow but unreliable DSL.[/QUOTE]

It worked for me (for anyone clicking on that original link, just note that you will get an assignment as a result). :smile:

LaurV 2019-05-20 03:15

Worked for me too. I got an assignment ([m]91338227[/m]) which I am going to work on. I removed the link from original post however. Curious people will click on it, and this seems to me a way to trick people in getting unwanted assignments, which they will not work, and will stay there and clutter the DB.

(No, it didn't trick me, because I read Madpoo's post first, before clicking, and I needed one new assignment, but it would for sure do trick me, if not for Madpoo's post, because I am a curious monkey!)

kriesel 2019-05-20 14:16

No trick intended. The point was, I went for a manual assignment, and instead of it working as usual, returning one or more assignments with aid(s), it returned an incomplete page as shown in the attached screen capture image, which might have been due to the maintenance going on at the time. It worked for me before that, and after that. It's not at all surprising that it would work days after the maintenance was completed; that [B]is[/B] the intended result of maintenance!

Madpoo 2019-05-21 04:14

[QUOTE=kriesel;517259]No trick intended. The point was, I went for a manual assignment, and instead of it working as usual, returning one or more assignments with aid(s), it returned an incomplete page as shown in the attached screen capture image, which might have been due to the maintenance going on at the time. It worked for me before that, and after that. It's not at all surprising that it would work days after the maintenance was completed; that [B]is[/B] the intended result of maintenance![/QUOTE]

Just a little behind-the-scenes...

When I'm going to do any maintenance like installing the Microsoft patches, my first step is to stop the web services before installing anything. There's usually a very good chance the patches are going to stop/start web services, especially when doing .NET framework stuff, so I just avoid it and stop it altogether.

I suppose if the timing is just right, I could stop the web services and something that was "in progress" could get killed off prematurely.

Typically though, when you stop the web server, it lets any outstanding requests complete first. The only time it would kill anything is if it took longer than the timeout period to complete (something like 30 seconds... I forget). I don't always pay attention to how long it takes to stop, but it's entirely possible.

Madpoo 2019-09-13 17:44

Server maintenance - 2019-09-13 @ 19:00 UTC
 
The server will be getting patches applied at 7 PM UTC (about 75 minutes from this post).

The updates will take ~ 30 minutes to complete during which time the server will be unavailable (both the website and the API used by the Prime95 clients).

I'll update this message when complete.

petrw1 2019-09-13 19:03

Brave man doing this on Friday the 13th during a full moon.

Madpoo 2019-09-13 19:34

[QUOTE=petrw1;525781]Brave man doing this on Friday the 13th during a full moon.[/QUOTE]

LOL...

All done, at any rate. :smile:

LaurV 2019-09-14 03:47

Haha, true story, my daughter was invited to a b'day party at some restaurant, girlie stuff, they were 16 invited, but 3 didn't show up, and during the party, between fun and drinks (not only non-alcoholic, as I understood) they realized that they were 13 and it was Friday 13, and she said it was a bit creepy for few minutes (then they switched to other funny things). She told us the story on skype, I mean like a fun story, and my reaction, of an old, grumpy hard atheist, not believing in such idiosyncrasies, was "interesting how far the coincidence can go... is the restaurant also called 13 cows or something"?

I don't think she'll talk to me next week... :razz:

kriesel 2019-11-11 17:12

login timeout during entering manual results
 
I've only had this happen once, recently, and didn't lose any data, but it's possible to do so if one is less cautious than I was. Sequence was this:
1) Note the manual result submission page shows me logged in
2) gather manual results from most of my systems, paste into the one manual result submission page
3) click submit
4) Server responds with a message that my login expired during the gathering, and tells me to log in before submitting results, after tossing my just-submitted results into its very own trashcan.

Maybe that behavior could be modified to preserve the submitted results, in a way that's easily available to the end user to resubmit after the login-again? (As in copy these results that did not go through, then login and try again by pasting them in.)

I was able to go back and gather them again, from the individual gpu results files, because I don't delete them ever, and they get annotated by the gathering program as to date & time of gathering to one location per system. The gathering per system had already gotten cleared, so the individual results files served as scattered numerous fragmentary backups.

kriesel 2019-11-11 17:23

Rounding
 
1 Attachment(s)
The upper limit of the manual gpu assignment page's 100Mdigit TF range could use some rounding. I don't think that value shown is factorable to any integers > 1. [url]https://www.mersenne.org/manual_gpu_assignment/[/url]

Madpoo 2019-11-17 18:45

[QUOTE=kriesel;530300]The upper limit of the manual gpu assignment page's 100Mdigit TF range could use some rounding. I don't think that value shown is factorable to any integers > 1. [url]https://www.mersenne.org/manual_gpu_assignment/[/url][/QUOTE]

Thanks for mentioning that. That's pretty weird. I may check on that page later and see how it's calculating that and switch it to round up or whatever.

Madpoo 2019-11-17 22:32

[QUOTE=Madpoo;530848]Thanks for mentioning that. That's pretty weird. I may check on that page later and see how it's calculating that and switch it to round up or whatever.[/QUOTE]

Fixed. It's doing a query that gets an average value and the result was a FLOAT. I just cast to an INT and all is well. :smile:

Madpoo 2019-12-08 23:05

FYI - New SSL cert on the web server
 
Just as an FYI, the server SSL certificate got updated today. Should be working fine, but I mention it here in case someone has a problem, at least they'll know what changed.

Madpoo 2019-12-13 21:26

Maintenance window - 2019-12-13 @ 21:45 UTC
 
The Primenet server will be taken down briefly for some maintenance and updates at 21:45 UTC (about 20 minutes from now).

I anticipate the downtime to last around 25 minutes - hopefully sooner.

Madpoo 2019-12-13 23:01

[QUOTE=Madpoo;532852]The Primenet server will be taken down briefly for some maintenance and updates at 21:45 UTC (about 20 minutes from now).

I anticipate the downtime to last around 25 minutes - hopefully sooner.[/QUOTE]

EDIT: Change of plan - got delayed starting so it'll be starting now @ 23:00 UTC

EDIT: Back online, good to go.

Madpoo 2020-02-02 05:07

Server maintenance - 2020-02-02 @ 05:15 UTC
 
I'm going to do a quick little maintenance on the server - apply some updates and reboot.

I should have thought ahead and planned it for 2:02 AM on 2020-02-02 (or even 02:20 to keep the palindrome thing going). Oh well.

Should only be down for approx 10 minutes.

EDIT: Took longer, but we're back in business.

Madpoo 2020-03-28 05:41

Server maintenance - 2020-03-28 @ 5:45 AM UTC
 
I'm going to do some routine updates on the Primenet server in about 5 minutes, at 5:45 AM UTC

Downtime should be in the range of 20-25 minutes, hopefully less.

I'll update when it's all done and we're operational again.

UPDATE: All done - took a little longer during the reboot phase, but it's back up and running now.

Madpoo 2020-05-24 05:24

Server update time
 
It's a good time for me to do some routine maintenance on the server (updates, some light housekeeping, etc).

I expect the downtime to start in about 7 minutes @ 5:30 AM UTC (10:30 PM PDT for me on the US west coast).

I'm expecting the downtime to last about 30 minutes since I have a few extra updates I've put off before, but figure it's time to do it.

Madpoo 2020-05-24 06:15

[QUOTE=Madpoo;546336]It's a good time for me to do some routine maintenance on the server (updates, some light housekeeping, etc).

I expect the downtime to start in about 7 minutes @ 5:30 AM UTC (10:30 PM PDT for me on the US west coast).

I'm expecting the downtime to last about 30 minutes since I have a few extra updates I've put off before, but figure it's time to do it.[/QUOTE]

All done. :smile:

Madpoo 2020-07-16 19:34

Offline for maintenance - 2020-07-16 @ 19:40 UTC
 
I'm about to take the Primenet server offline for standard software patching. About 7 minutes from right now at 19:40 UTC

Downtime should be the usual of about 20-30 minutes.

EDIT: Late start... beginning now @ 20:00 UTC

Madpoo 2020-07-18 18:25

Quick maintenance - Primenet offline for about 2 minutes
 
I'm taking Primenet offline real quick for a brief maintenance task related to some updates we're working on. Should be back online pretty quick, maybe 2-3 minutes.

Madpoo 2020-12-31 16:53

Offline for standard maintenance - 2020-12-31 @ 16:55 UTC
 
Hi all - the Primenet server will be going offline for about 30 minutes for some standard maintenance. It's been a while since the last time so the year-end seemed like as good a time as any to catch up.

EDIT: Still going... be patient. :smile: There were a number of updates waiting, so it's taking a bit longer than expected. Multiple reboots involved.

Madpoo 2020-12-31 18:08

[QUOTE=Madpoo;567858]Hi all - the Primenet server will be going offline for about 30 minutes for some standard maintenance. It's been a while since the last time so the year-end seemed like as good a time as any to catch up.

EDIT: Still going... be patient. :smile: There were a number of updates waiting, so it's taking a bit longer than expected. Multiple reboots involved.[/QUOTE]

All done.

Madpoo 2021-01-25 02:02

Quick reboot - 2021-01-25 @ 02:05 UTC
 
Quick reboot of the server... should only be a few minutes. :smile:

EDIT: Good to go.


All times are UTC. The time now is 03:38.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.