- luv-main - luv.asn.au

Re: Fwd: Postfix "trickle" list delivery
by Petros 04 Jun '13

04 Jun '13

From "Russell Coker" <russell(a)coker.com.au> > http://aws.amazon.com/ec2/#pricing > > A "small" EC2 instance costs 6 cents per hour or $43.80 per average month. > Plus costs for bandwidth and storage. A "micro" instance costs 2 cents per > hour or $14.60 per average month. .. > EC2 is really good for dynamically scalable systems and for situations where > you need bulk computing off-peak (as demonstrated in a LUV talk some > time ago). > It's not particularly good for running a single server. As said, it is an announcement server which is "fed" by us maybe once a month. The idea was to let it run only when needed. I wasn't aware of SMTP traffic restrictions. I actually looked out for it but could not find it mentioned anywhere. Besides, for me it was a trial to figure out how to deal with Amazon EC2. Well, actually, fears confirmed: you never know exactly what you get, you cannot rely on it and you don't get what you want;-) Thanks for the other ideas as Linode or Hetzner. I have a think about it. Regards Peter

2 1

Re: Postfix "trickle" list delivery
by Petros 03 Jun '13

03 Jun '13

Hi Russell, thanks for the answer. From: "Russell Coker" <russell(a)coker.com.au> > On Mon, 3 Jun 2013, Petros <Petros.Listig(a)fdrive.com.au> wrote: >> If I send many e-mails at the same time, I have a high number of >> re-delivery attempts because initially connections time out. >> >> I am playing with some parameters in main.cf but haven't been very >> successful yet. >> >> At the moment my main.cf says: >> >> smtp_destination_rate_delay = 120s >> smtp_destination_concurrency_limit = 1 >> initial_destination_concurrency = 1 >> default_process_limit = 5 > > http://www.postfix.org/postconf.5.html#smtp_destination_concurrency_limit > > smtp_destination_concurrency_limit is per recipient. So it will reduce the > number of connections when sending to gmail and yahoo but not much else. Yes. I would have liked one that would limit the number of outgoing SMTP connection globally but I could not find a parameter for it. > From the same URL the default_process_limit is the number of processes that > the master process spawns, this will affect the DEFAULT for inbound > connections but not much else. Yep, seems so. Although it could be used to delay transfer from mailman to postfix, I thought. At the moment the recipient list is limited to 50 per connection (so mailman divides the list in chunks when sending mail). If I send it to 1 (default_process_limit = 1) it would restrict the number of recipients to 50 at the time, I think. I just don't know when mailman is sending the next chunk then (after one connection is closed?) >> BTW: The server is in the Amazon EC2 cloud but I don't think it >> matters. Still, if I try a telnet on port 25, the connection opens >> faster from my office (ADSL) than from the Amazon VM. > > Is there a reverse DNS issue? Well, the reverse lookup resolves to ec2-...us-west-2.compute.amazonaws.com not to the used sender address because it is a dynamic address. At the moment I update it on internal DNS (I don't control the external) so if mail arrives at the company server it can redirect list mail to the Amazon computer. > EC2 should be fast enough. How many users are on the list? Ca. 500 users. It is an announcement lists only. At the moment, I still have 56 mails left, after ca. 3 hours. It took ca. an hour to deliver the first 250. The main reason is "Connection time out", according to mailq. The mail itself is a PDF with less than one 1 Mbyte. I am surprised by the slowness. I wonder whether Amazon is throttling SMTP traffic. We have our mail server inhouse, and having the list server inhouse is clogging the ADSL connection. So I try Amazon, I thought. We do not own rack space otherwise I would have done it in a datacentre. Regards Peter

1 0

Postfix "trickle" list delivery
by Petros 03 Jun '13

03 Jun '13

Hi all, I have a mailman server with postfix. If I send many e-mails at the same time, I have a high number of re-delivery attempts because initially connections time out. I am playing with some parameters in main.cf but haven't been very successful yet. At the moment my main.cf says: smtp_destination_rate_delay = 120s smtp_destination_concurrency_limit = 1 initial_destination_concurrency = 1 default_process_limit = 5 A delayed delivery is acceptable so I just try to slow down to avoid timeouts. http://www.postfix.org/QSHAPE_README.html says: "When a host with lots of deferred mail is down for some time, it is possible for the entire deferred queue to reach its retry time simultaneously. This can lead to a very full active queue once the host comes back up. The phenomenon can repeat approximately every maximal_backoff_time seconds if the messages are again deferred after a brief burst of congestion. Perhaps, a future Postfix release will add a random offset to the retry time (or use a combination of strategies) to reduce the odds of repeated complete deferred queue flushes." I guess this feature would help.. Anyway, do you any more ideas how to improve the delivery? BTW: The server is in the Amazon EC2 cloud but I don't think it matters. Still, if I try a telnet on port 25, the connection opens faster from my office (ADSL) than from the Amazon VM. Thanks for all ideas Peter P.S. I am fiddling with it for so long, I typed luv-main.cf first;-)

2 1

calendar program?
by Craig Sanders 01 Jun '13

01 Jun '13

iceowl has finally stopped working in debian sid (mozilla sunbird was discontinued upstream a long time ago). does anyone know of a simple stand-alone calendar program that can sync with google calendar? mostly, i just want something where i can enter appointments etc into the calendar on my desktop and have them turn up on my phone and tablet. and vice-versa. things that aren't suitable: 1. I don't want to use the web-based google calendar because i don't want to leave a browser logged in to google apps all the time. 2. I already know about icedove and the lightning extension. As a last resort, it will do (and it is almost certainly what i'll use until i find something else) but i just want a standalone calendar program - I don't want it messing with my mail too. iceowl was the standalone calendar-only version of icedove+lightning. it worked exactly as i wanted it to, and the UI was more than adequate for my purposes. 3. Anything that depends on gnome shell or requires a specific desktop environment. I use a variety of apps from gnome, xfce, kde, and others but i have no intention of becoming dependant on any one of them, otherwise i'd have to completely change my entire desktop environment after every major release when gnome/kde/whoever perpetrates their next we-know-what's-best-for-you idiocy. craig -- craig sanders <cas(a)taz.net.au>

3 4

Re: Screen Res. Was alternate OS voids warranty
by mjones11 01 Jun '13

01 Jun '13

-------- Original message -------- From: Russell Coker <russell(a)coker.com.au> Date: 30/05/2013 11:25 (GMT+10:00) To: luv-main(a)luv.asn.au Cc: muzza <mjones11(a)tpg.com.au> Subject: Re: Screen Res. Was alternate OS voids warranty On Wed, 29 May 2013, muzza <mjones11(a)tpg.com.au> wrote: > woefully pitiful especially considering that some MS Windows configuration > windows often have the OK and cancel buttons below the bottom of the > screen and you have to tab to them and hope you know which one you > selected. This is even worse in linux for some reason. OK I digress - the > point is why are they Hold down the ALT key, then you can click the mouse anywhere in a Window to drag it around. This allows you to get access to the bottom of a window which is larger than the screen resolution. Yes this works and it is possible to get around those problems in windows as well. Its just a pita that developers don't consider and testers are slack

2 2

Re: fast fail on specific entries not defined in /etc/hosts
by Petros 01 Jun '13

01 Jun '13

From: andrewv(a)melbpc.org.au > This one sounds like the most viable option for me. I can run a DNS > resolver on the same host as the application is running on, and configure > it just for this application, and change the DNS search order to consult > this local DNS service first (and then the ones it currently uses). I can > use multiple domains if needed (not just company.com) You could use the local DNS server and configure this to forward everything else (not needing "special treatment") to the "normal" DNS servers. /etc/resolv.conf would need 127.0.0.1 only. Regards Peter

1 0

Re: fast fail on specific entries not defined in /etc/hosts
by andrewv＠melbpc.org.au 31 May '13

31 May '13

Hi, Thanks for the feedback. See my responses below: > andrewv(a)melbpc.org.au writes: > > > Assume I have (at most), all of these entries in /etc/hosts > > > > custom1.company.com nnn.nnn.nnn.nnn > > custom2.company.com nnn.nnn.nnn.nnn > > default.company.com nnn.nnn.nnn.nnn > > That won't work. It's > > <address> <canonical name> [<alias> ...] > Sorry, you're right - my typo :) > > where nnn is some specific IPv4 address number > > > > I want to lookup the hosts in say this order: > > > > custom2.company.com > > custom1.company.com > > default.company.com > > It doesn't work that way, sorry. > I wasn't clear enough - I meant the application will specifically attempt to look up custom2.company.com, and if not found, will try to look up custom1.company.com, then if not found, will lastly try default.company.com. I wasn't trying to say that this search order had anything to do with how the lookup/resolving will be done for me by the lookup service. > > If the custom2 entry is not found, I want to try custom1, and if not ok, > > then I want to try default. > > You're talking about reverse resolution (IP to FQDN)? > No, I mean this is what my application will explicitly attempt, in that order. > > Importantly, I do _not_ want _any_ DNS lookup to be performed for > > _these_ _specific_ host lookups, because I want an immediate failure > > if the entry is not defined in /etc/hosts. > > I don't think you can have that, at least not on a per-host basis. > ok > The way different name services (such as flat files and DNS) are used is > governed by nsswitch on eglibc/glibc-using (and most other) systems. > > You can probably add > > 1.2.3.4 canthappen.invalid > > And if nsswitch has "hosts: files dns" then it will hit that first, and > not try DNS. You can't have an absence in /etc/hosts -- hosts(5) has no > support for the equivalent of an authoritative NX RR. > ok > > Note: I want an _immediate_ failure initially, _and_ for every lookup > > thereafter - an initial DNS lookup that fails after a DNS lookup > > timeout, and then perhaps caches that failure result is no good, > > because every lookup must respond immediately (and of course > > /etc/hosts and DNS service entries can change at any time so even > > then, cacheing is not useful). > > You can't have this with hosts(5). > ok > Your best bet is probably to point resolv.conf at a local DNS resolver > that is configured to return spoof results for particular domains > (including, obviously, the C.B.A.in-addr.arpa for the PTR RR). > This one sounds like the most viable option for me. I can run a DNS resolver on the same host as the application is running on, and configure it just for this application, and change the DNS search order to consult this local DNS service first (and then the ones it currently uses). I can use multiple domains if needed (not just company.com) > You could, alternatively, disable DNS altogether and ensure that all > hosts you need to reach are hard-coded in /etc/hosts. > Not an option, because many other lookups are performed aside from those servicing this application. > > Assume that I am using software that I cannot change, and so workarounds > > like actually reading the /etc/hosts file to see which entries exist, > > cannot be wired into the software. > > LD_PRELOAD might still be a viable option, though I don't recommend it. > Considered that one, I see it as last resort stuff, but yes, viable if nothing else hits the mark. > > I am thinking of something functionally like a like /etc/disable_dns_lookup > > that I can also copy all 3 of the above entries to, and linux will _never_ > > try DNS if the entries appear in /etc/disable_dns_lookup. > > AFAIK no such feature exists. > > This all sounds like an xyproblem; what is the ACTUAL problem you're > trying to solve? > I want to be able to use the configuration entered in /etc/hosts, as the sole data driven means of configuring host selection order, assuming the hosts are always going to be on an internal network, with well known addresses. > _______________________________________________ > luv-main mailing list > luv-main(a)luv.asn.au > http://lists.luv.asn.au/listinfo/luv-main Message sent using MelbPC WebMail Server

2 1

fast fail on specific entries not defined in /etc/hosts
by andrewv＠melbpc.org.au 31 May '13

31 May '13

Hi, I wanted to know if the following was possible on a linux (in particular) box: Assume I have (at most), all of these entries in /etc/hosts custom1.company.com nnn.nnn.nnn.nnn custom2.company.com nnn.nnn.nnn.nnn default.company.com nnn.nnn.nnn.nnn where nnn is some specific IPv4 address number I want to lookup the hosts in say this order: custom2.company.com custom1.company.com default.company.com Now sometimes, I may only have the default entry, and sometimes I may have the custom1 and / or custom2 entries also in /etc/hosts. So, any combination from 0 to all 3 entries is possible at any point in time. This configuration would not change often, but any combination of entries is possible. If the custom2 entry is not found, I want to try custom1, and if not ok, then I want to try default. Importantly, I do _not_ want _any_ DNS lookup to be performed for _these_ _specific_ host lookups, because I want an immediate failure if the entry is not defined in /etc/hosts. Note: I want an _immediate_ failure initially, _and_ for every lookup thereafter - an initial DNS lookup that fails after a DNS lookup timeout, and then perhaps caches that failure result is no good, because every lookup must respond immediately (and of course /etc/hosts and DNS service entries can change at any time so even then, cacheing is not useful). Assume that I am using software that I cannot change, and so workarounds like actually reading the /etc/hosts file to see which entries exist, cannot be wired into the software. I am thinking of something functionally like a like /etc/disable_dns_lookup that I can also copy all 3 of the above entries to, and linux will _never_ try DNS if the entries appear in /etc/disable_dns_lookup. Any solutions / ideas ? Thanks Andrew Message sent using MelbPC WebMail Server

2 1

Re: Screen Res. Was alternate OS voids warranty
by mjones11 31 May '13

31 May '13

-------- Original message -------- From: Toby Corkindale <toby.corkindale(a)strategicdata.com.au> Date: 30/05/2013 10:54 (GMT+10:00) To: luv-main(a)luv.asn.au Subject: Re: Screen Res. Was alternate OS voids warranty On 29/05/13 21:16, muzza wrote: > In the past yoiu could get decent res:- 1280x1024 or above and quite commonly > 1600x1200 and at 2ms response times. However it seems that some idiot > somewhere decided that we only watch videos on our computers so we only need > 1920x1080 at 5ms response times, and for crappy lappies only 1366x768. 27" Dell Ultra Sharp ultra U2713H - 2560x1440 - $799 27" X-Star DP2710 (Samsung panel) - 2560x1440 - $295 I would consider the ultra Sharp to be expensive. Never heard of the X-Star so can't comment Murray

2 1

Re: Screen Res. Was alternate OS voids warranty
by zlinw＠mcmedia.com.au 30 May '13

30 May '13

On Thu, 30 May 2013 10:54:03 Toby Corkindale wrote, > In the past you could get decent res:-1280x1024 or above and quite commonly > 1600x1200 and at 2ms response times. However it seems that some idiot > somewhere decided that we only watch videos on our computers so we onlyneed > 1920x1080 at 5ms response times, and for crappy lappies only 1366x768. 27" Dell UltraSharp U2713H - 2560x1440 - $799 27" X-Star DP2710 (Samsung panel) - 2560x1440 - $295 There are options for higher-res screens; they've been around for a while too. They're just not as common. Lindsay Replies, 16x10 monitors (business class) are certainly easier to get now than a year or 2 back. Particularly 2560x1600. I only use 16x10 monitors as the others simply do not have enough height. I have two 2560x1600, I purchased one a Samsung 305t some years ago at great expense. I liked it ___SO___ much I purchased a HP ZR30w a few months ago, it was around $1300 from a shop in Shepparton. The 30 inch monitors were originally purchased for graphics and photo work. But they are ___absolutely___ made for train sims, Really!!!! The 2560x1440's do not have enough space for a decent cabview, most loco cabs being more or less arranged in a vertical fashion. (the 2560x1600 do not really have enough space either but the don't make 2560x1920 monitors and probably never will). Hmmmmmmmmmmmmmm two 2560x1440's vertical side by side may be interesting. I know someone who has gone for three 1920x1080's vertically side by side specficly for trainsim work. Lindsay

1 0