#arpnetworks 2016-09-29,Thu

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
***pyvpx is now known as sysvival
sysvival is now known as OMGKitteh
OMGKitteh is now known as Guest88910
Guest88910 is now known as pyvpx
pyvpx is now known as Guest6369
Guest6369 is now known as _click
_click is now known as pyvpx
[05:17]
........... (idle for 53mn)
nznzx2 has joined #arpnetworks [06:19]
........................................... (idle for 3h30mn)
brycec<Obligatory "You don't, you use a sensible web server like Nginx" response/> :p [09:49]
mhoranI'm always surprised when I hear people still using apache.
and if you're already using php_fpm, why even use apache!
[09:52]
...... (idle for 25mn)
brycecPrecisely.
Only thing I find Apache useful for: home directories, eg. blah.com/~user/
[10:17]
***tabthorpe has joined #arpnetworks [10:24]
.... (idle for 18mn)
mercutiolots of of people use it for htaccess
adding htaccess and ~ support would probably get a lot of people off of it :)
oh the htaccess support would have to support apache rewrite rules too
[10:42]
.... (idle for 15mn)
brycecYeah, though I can see why most webservers probably haven't picked up those features. They're very "multiuser" features (as in for systems with many users, where the users need some modicum of server configuration control) which seems to have largely fallen by the wayside these days. Either the user is the sysadmin and experienced, can modify the server configuration as necessary, host their files
wherever, or the user signs up for a {Tumblr,Wordpress,whatever} account.
[10:58]
mercutiowell installation instructions are often easier to follow for apache
with the .htaccess
[11:00]
mike-burnsdoas pkg_add nginx && doas vi /etc/nginx/nginx.conf && doas /etc/rc.d/nginx start [11:01]
***Seji has quit IRC (Ping timeout: 264 seconds) [11:03]
..... (idle for 21mn)
brycecPretty sure mercutio was referring to the installation instructions given with projects like Wordpress or Drupal :P Not the server itself. [11:24]
mercutioyeah i was [11:25]
......... (idle for 42mn)
nathaniSome hosts use nginx as front end proxy with Apache backend
So htaccess works
[12:07]
........ (idle for 35mn)
***joepie91_ has quit IRC (Ping timeout: 252 seconds) [12:43]
joepie91_ has joined #arpnetworks [12:54]
mnathanihttp://forums.hostgator.com/nginx-t305821.html
HTTP headers report nginx 1.4.7, but phpinfo() reports Apache.
[12:56]
........... (idle for 54mn)
***joepie91_ has quit IRC (Ping timeout: 252 seconds) [13:51]
joepie91_ has joined #arpnetworks
Seji has joined #arpnetworks
[14:05]
..... (idle for 21mn)
nathani has quit IRC (Quit: WeeChat 1.4)
nathani has joined #arpnetworks
Seji has quit IRC (Ping timeout: 272 seconds)
[14:30]
..................... (idle for 1h43mn)
up_the_ironsWould anyone be interested in simple hosted MySQL, PostgreSQL, Redis, MongoDB, etc... service that was super high performance because it's backed my the massive parallel read/writes afforded to us by a Ceph cluster? Thinking it would be a value-add to our VPS / dedicated servers and not even necessarily a thing we'd charge for (but a paid tier would probably exist for those that wanted more space than
the free value-added one would allow)
Just putting that out there...
backed *by
I would tweet that if it would fit LOL...
[16:16]
........ (idle for 39mn)
mercutioyou could just ask about hosted databases on twitter [16:57]
brycecup_the_irons: As it happens, my employer would be interested in hosted Postgres. We were planning (tentative, few months out) to use Redshift/DynamoDB/etc for a "hosted, scalable and performant database" so if ARP had an offering, we'd be interested. [17:05]
........... (idle for 52mn)
***Nahual has joined #arpnetworks [17:57]
...... (idle for 26mn)
up_the_ironsbrycec: so single hosted Postgres DB, any special requirements?
mercutio: how would you word it?
[18:23]
mercutioWe're exploring interest for Hosted Databases. Who would be interested in such?
or something like taht
it's a bit weird having proper nouns before you have a name :)
maybe hosted MySQL, PostgreSQL, Redis, etc..
[18:24]
up_the_ironsgot it :)
tweet sent :)
[18:25]
mercutiogizmoguy: your power monitoring looks cool
yeah i see it, and i saw gizmoguy's power graphs too :)
[18:28]
gizmoguyyeah it's quite cool!
http://power.sla.ac
[18:32]
mercutioi'm not sure how to hook it up
so you can tap into the mains power somehow at the base?
[18:32]
gizmoguywe're counting the pulses coming from the smart meter [18:33]
mercutiooh so you need a smart meter? [18:33]
gizmoguysomething that generates pulses [18:33]
mercutiodo normal power meters generate pulses? [18:34]
gizmoguyhttps://openenergymonitor.org/emon/opticalpulsesensor
this is the sensor we're using
mercutio: you'll have to go have a look at your power meter
[18:34]
mercutioahh and then google it? [18:34]
gizmoguyif it's a digital (non-smart) meter it should still do pulses [18:34]
mercutiooh cool [18:34]
gizmoguybasically if it has a flashing red light on it
you should be good
[18:34]
mercutiohah shit
it's raining hard
that caught me by surprise
i'll wait 10 minutes :)
[18:35]
gizmoguyhahaa [18:35]
mercutioi suspect it is digital non smart
but who knows
they've actually wanted to install smart meter for ages
but we told them weren't interested
if smart meters had graphs like this i'd go for them
[18:36]
gizmoguyso they do
but from talking to my friends who work for power companies they usually can't be bothered to store the fine grained data
so genesis will only store hourly data for example
even though they get it 5-minutely I think
[18:39]
mercutioyeah it's a lot of data
oh they get every 5 minuntes
[18:39]
gizmoguyon the smart zigbee ones it's something like that
since we're counting pulses we get it instantly, but we're binning by minute
[18:40]
mercutiocan you do more than once a minute? [18:40]
gizmoguyyip
we could move down to 30 second or 10 second
[18:41]
mercutiodamn
it stopped raining :)
[18:41]
gizmoguyI'm not sure how accurate the RTC on our sensors are though [18:41]
mercutioyeah it's not even digital
s200.31
funny how you can google anything now
and find youtube videos :)
[18:44]
gizmoguyYeah so that will be fun to monitor
Your pulse is that spinning wheel
[18:50]
mercutiowouldn't it be easier to monitor one jackpoint or something? [18:50]
gizmoguyYou can get the clamp on monitors
it'll be quite expensive though
just get a smart meter
[18:51]
mercutioahh maybe when i move
nearly everywhere has them anyway
so how hard is it monitor 10amp circuits etc?
mercutio hasn't even started monitoring his ups yet
[18:52]
gizmoguyhttps://www.sparkfun.com/products/11005
looks cheap
[18:58]
mercutiooh wow
damn
now i'm getting tempted :)
wow shipping prices aer ok there
[18:59]
***Nahual has quit IRC (Ping timeout: 272 seconds) [19:07]
.......... (idle for 49mn)
brycecThat it not suck :P In all seriousness though, nothing "special" I don't think. It would be nice if the traffic stayed on-net (privacy concerns, and avoiding using our bandwidth quota), and highly ("guaranteed") available. Or were you looking for a different "special requirements" answer? 18:23:16 @up_the_irons | brycec: so single hosted Postgres DB, any special requirements? [19:56]
up_the_ironsbrycec: nope, that's perfect, thanks :)
"on-net" is interesting... I'm thinking about how that would be done
[19:57]
mnathaniwithout having to manage rfc1918 space amongst all customers using the service [20:02]
mercutioit could be like the backup service [20:08]
brycec^
I mean, I imagine the connection would be TLS-wrapped in the first place, but keeping it on-net is an extra bit of warm-fuzzy security.
[20:08]
mercutioi'm a little paranoid about mysql's security
so on-net only seems preferable to me
[20:10]
brycecwell fuck MySQL :P (just my personal preference) [20:10]
mercutiobrycec: you know people will want it [20:11]
brycecPeople don't know what they want... [20:11]
mercutiohttps://www.cvedetails.com/vulnerability-list/vendor_id-185/product_id-316/Mysql-Mysql.html [20:11]
mnathanidefine on-net?
at the same datacenter?
[20:12]
mercutiolocal network
although on-net doesn't mean "private"
[20:12]
mnathaniso no routers
layer 2 network?
[20:12]
mercutioso it could be over normal internet ip addresses but be acl'ed, and still be on-net [20:12]
brycecNot crossing/touching the public router.
Though I would accept VPN'd traffic
[20:12]
mercutiooh brycec has a different idea of on-net to me :) [20:13]
mnathaniwouldnt TLS or VPN be a huge performance hit for database traffic? [20:13]
brycec(Ooh there's a thought, use VPN for extra added ACL to the database stuff) [20:13]
mercutiotls is no concern at all now days [20:13]
brycecTLS/VPN a huge performance hit? Next you'll be warning me about Y2K. [20:14]
mnathaniits a question, I dont really know [20:14]
brycecLike mercutio, modern computers handle that like they handle blinking an LED, no significant overhead. [20:14]
mercutiobrycec: it was a huge overhead in apache [20:15]
brycecHow do you mean? (As someone that works from home 100%, all my traffic including to an Apache server at work is VPN'd with no noticeable issues.)
Like, are we talking adding a millisecond? Seconds?
[20:16]
mercutioit used a lot of memory, back when memory was epensive
the overhead of adding ssl onto sites increased memory requirements a lot
[20:17]
brycecOhhh you mean Apache as the SSL endpoint. [20:17]
mercutioyeah [20:17]
brycec(I was mentally stuck on VPNs) [20:17]
mercutioopenvpn isn't as efficient as it could be
but for most uses straight tls is fine
[20:17]
brycec(And TLSv1.3 is looking promising for reducing latency/improving connection times) [20:18]
mercutioyeah i read about that from microsot implementing tcp fast open
tls quickstart kind of failed before
[20:19]
brycecCloudFlare just announced TLSv1.3 terminations/endpoints.
https://blog.cloudflare.com/introducing-tls-1-3/
[20:19]
mercutiooh wow
is chrome doing it yet
[20:20]
brycec(no idea) [20:20]
mercutioi find cloudflare kind of slow :( [20:20]
brycecFirefox 49 does, according to Googling [20:21]
mercutiobut they're definitely doing a lot of innovative interesting things [20:21]
brycec"Experimental, disabled by default" [20:21]
mercutioand getting a lot of market share [20:21]
brycechttps://en.wikipedia.org/wiki/Template:TLS/SSL_support_history_of_web_browsers [20:21]
BryceBotTemplate:TLS/SSL support history of web browsers :: Notes ➔ Back to article "Transport Layer Security" References... [20:21]
mercutiofirefox 49 hmm [20:22]
.... (idle for 18mn)
up_the_ironsI would imagine it being something like the backup service, but probably not require an extra NIC... [20:40]
mnathaniI thought ceph was designed for storage. Does it handle highly available databases too? [20:40]
up_the_ironsI mean, your VM NIC is already VLAN tagged... on the "other" end (hosted DB box), it would terminate your VLAN, among other customers who also have the service
mnathani: no it doesn't, but your DBs can use an rbd volume for their storage
[20:40]
mercutioit has a generic object store [20:41]
up_the_ironsAn rbd volume from Ceph can be faster than a single disk, b/c it stripes the volume over the cluster [20:41]
mnathaniso the compute portion of the database would be tied to an individual ceph host
or other db host that uses ceph for storage
[20:41]
mercutioyeah [20:42]
up_the_irons"other db host that uses ceph for storage" is the right answer [20:42]
mnathanithat makes it high performance, not highly available [20:43]
up_the_ironsand *that* DB host can be a very large VM too, which has the ability to be live migrated around the cluster, if needed or in the case of hardware failure [20:43]
mercutioit could migrate between hosts
it makes it HA in normal HA speak
[20:43]
up_the_ironsIt would be more highly available than a DB on a single host [20:43]
mercutiobut not active/active ha [20:43]
up_the_ironsb/c that DB would be stuck to that host [20:43]
mnathanihigh available to me means it can survive a node going down [20:44]
mercutiosurvive means how much downtime mnathani ? [20:44]
mnathaninone
it would have to be behind a load balancer
[20:44]
mercutioso HA means to you active/active [20:44]
mnathaniand active active
yea
[20:44]
up_the_ironsload balancer ain't a bad idea [20:45]
mercutiohaha
https://www.google.co.nz/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwii28LnlrbPAhVJ6oMKHdb2B5MQFggcMAA&url=https%3A%2F%2Fwww.ibm.com%2Fdeveloperworks%2Fcommunity%2Fblogs%2Fmhhaque%2Fentry%2Fconfigure_two_node_highly_available_cluster_using_kvm_fencing_on_rhel7&usg=AFQjCNH21Ebq2hUmT10-NWYh95P5BNeWfA&sig2=L3t8gg8JNEBIBM3snE7LWA
oh i thought it was a normal url
but no it's a tracking url
[20:45]
mnathaniI get Sorry! developerWorks is offline [20:46]
mercutioexactly
but look at the url
[20:46]
mnathaniyea it is funny
even active/passive would be highly available
[20:46]
mercutiowell with ceph active/passive is easy
if host goes down can start on another host
[20:47]
mnathanias long as the passive node promotes itself to active upon failure
would you track cpu usage by the database?
like bill for more recurring complex queries
or perhaps by i/o bandwidth
[20:47]
up_the_ironsi don't think we'd go that granular [20:49]
mnathanifair use policy applies I guess
should not impact other users
would you require ARP VPS or Dedicated to access the database, or allow external hosts as well
perhapas as secondary users inspecting the database and running routine queries, but not the primary user - that would be the ARP VPS or Dedicated box
[20:50]
up_the_ironssince access would come through account VLAN (most likely), I imagine it could be accessed through either a VPS or dedicated server
i mean, that would be a good way to do it
and not have complicated access rules
[20:55]
............................... (idle for 2h34mn)
***Seji has joined #arpnetworks [23:30]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)