I have now tried both, #memcache and #redis, to set single keys in cache. It seems both perform very poor compared to #in-progress caching.

So really no improvement if they are being used? And pipelining in redis won't help here much as I really have to "atomically" set/test/get key-value pairs.

So my in-progress cache as following seems to be the fastest:

function someCachedFooValue ($someValue) {
	if (!isset($cache[__FUNCTION__][$someValue])) {
		$cache[__FUNCTION__][$someValue] = doSomethingFooExpensive($someValue);
	}

	return $cache[__FUNCTION__][$someValue];
}

Here I want to cache the value from the expensive (long-taking) function doSomethingFooExpensive() if it is is not here.

This way seems to be the fastest way, sadly.

I have this on my server:

___________total       used       free     shared    buffers     cached
Mem:       7939312    6490588    1448724          0      66760     192064
-/+ buffers/cache:    6231764    1707548
Swap:     12485748     189256   12296492

Here are all cache settings:

# sysctl -a|grep cache
kernel.sched_domain.cpu0.domain0.cache_nice_tries = 1
kernel.sched_domain.cpu1.domain0.cache_nice_tries = 1
kernel.sched_domain.cpu2.domain0.cache_nice_tries = 1
kernel.sched_domain.cpu3.domain0.cache_nice_tries = 1
vm.drop_caches = 0
vm.vfs_cache_pressure = 100
fs.quota.cache_hits = 0
net.ipv4.rt_cache_rebuild_count = 4
fscache.object_max_active = 4
fscache.operation_max_active = 2

Da haben die #Krefeld #Pinguine mal ihre Homepage neu gemacht und sogar mit gezwungenem #SSL und was sieht man? Kein #RSS mehr moeglich.

Wieso? Wollen die nicht, dass sich deren Nachrichten einfach im Internet verteilen oder sollen die nur ueber einen zentralisierten Kanal erreichbar sein?

Und oh, #ssllabs gibt denen eine B-Note: TLS_DHE_RSA_WITH_AES_256_GCM_SHA384 (0x9f) DH 1024 bits FS WEAK z.B. gibt es. Das sollten die mal ueberarbeiten.

I was able to message to myself over an other XMPP server. But setting up #OTR did fail in endless reloads of the page causing OTR to be re-initialized and another session being started.

How can I prevent this? #Pidgin is showing them all and seems not to work then.

I wonder if the poller is really needed or if somehow the cron jobs can be executed without it? I had setup as it should be:

*/10 * * * * cd ${HOME}/friendica/htdocs/ && /usr/bin/php include/poller.php

If I let it runs, multiple php processes are forked eating up my CPU and slowing down the server. Currently it seems to work without it (and much faster).

I currently try to setup the XMPP plugin (jappixmini is not enabled). I have so far changed my ejabberd.cfg and made http-bind working (Apache2 config entry, really easy to setup).

Now I hang around with BOSH host (which address:port to enter?) and that the XMPP chat is not showing up. Any other things I need to change?

Ich glaube, der Poster im Forum hat recht: nur Panikmache bei Heise.de. Ich habe mal fix die besagte Dabei geprüft und root besitzt sie:

-rw-r----- 1 root root 4602 Jul 26 00:37 my.cnf

Also alles ruhig bleiben und nur kurz prüfen ...

I'm currently rescuing files from a partly broken hard drive where the very early sectors are unreadable. It is a USB hard-drive ("mobile") from a DJ. So I will find it "easy" to search for #MP3 files (he doesn't use OGG).

First I tried to normally mount it, which fails due to the broken sectors 0-9 (which holds the required data). Also a fsck.vfat didn't work. So I had to install #myresuce and and #magicrescue.

I did it better in screens so I can leave them unattended:

screen -dmS myrescue myrescue /dev/sdb1 sdb1.bin

I had to wait ~ 10 minutes to start the next one:

screen -dmS magicrescue magicrescue -d sdb1/ -r mp3-id3v2 sdb1.bin

/dev/sdb1 is the USB drive and of course I needed to create ~/sdb1/ to have found files being written there.

Now there are a lot MP3 files but all 40 MB large. I will forward them to the DJ so he has to cut them to the right size.

I just thought you might be interested in before going to a "professional" #disk #rescue company.

I currently face a #bug in #Payara that doesn't allow TABLE_PER_CLASS and I cannot change the database layout (as then the main program needs to be changed which involves a lot risky work).

I have filled out a bug report ...
github.com/payara/Payara/issue…

... But they don't want to handle it. I guess I'm not alone with this bug. I don't like such "payment model" where paying customers are priotized. I for my projects provide unpriotized support for any one, unpayed and in my free time:
My View - Mailer-Project Bug-Tracker
What do you think about such payment models? Good or bad?

2nd submit: better preview before test.

I currently get this when I try to deploy my project:

org.eclipse.persistence.exceptions.ValidationException Exception Description: The type [class java.time.DayOfWeek] for the attribute [endDay] on the entity class [class org.mxchange.jcontactsbusiness.opening_times.BusinessOpeningTimes] is not a valid type for an enumerated mapping. The attribute must be defined as a Java enum.. Please see server.log for more details.

I'm trying to fix this by writing a AttributeConverter<DayOfWeek, String>.

On Fri, 19 Jul 2002 a user called "burris" has committed the first files for a pre-project. Now, 2016 almost nothing happens to it anymore. The author #CameronDale has seem to given it up and I have almost no time to continue on.

If you search for it:
You find 2 projects, maybe indepently created.

So maybe this weekend or next weekend I try to merge them both into mine and notify them to merge back to have all efforts combined.

You can find mine here:
git.mxchange.org/?p=quix0rs-ap…

The mentioned message still comes when you have @Inheritance (strategy = InheritanceType.TABLE_PER_CLASS) in your abstract (not-persisted) with latest version of #Payara / #Glassfish which both contains #EclipseLink by default.

Strange that they did not yet fix it. And unfortunally migrating to e.g. #WilfFly (which uses other #JPA) seems not so easy as you may think.

Currently I have no idea how to fix it and sadly I need the strategy as this #JavaEE application is written for an existing database layout.

Someone may have to raise a support ticket mentioning that the #bug is still there.

Reposted with #public visibility. Sorry for double message.

I have now installed a better #SSL certificate issued by #LetsEncrypt and with the help of this #python script: github.com/diafygi/acme-tiny

Plus I have used a largely expanded #shell script:
social.mxchange.org/renew-cert…

But now my SSL sites got rated with B by #SSLLabs because of an incomplete chain. I think I need to download the CA certificate and attach it at mine?

I also have (not yet reloaded) added DH parameters (DH = Diffie-Hellmann?) to my certificate parameters, I have read it will improve some security.