Without restart or modification of config, nginx is sending incomplete responses sometimes. Today, I could reproduce well, but still don't know what's wrong or how to fix.

I put a 5MB file to the site's location/alias path that nginx is configured to use. There is no php5-fpm, modules etc. only nginx for serving static files.

When testing, there is no other user accessing the server, no http access except my test requests.

The download fails and Google Chrome shows in traffic log, that there were 2 requests, although I put the http url for the download only ONCE and there was no redirect or other 2nd request initiated by myself.

Same issue at server's access.log, too:

TEST1

[14/May/2014:14:06:53 +0200] "GET /dev/test_test.m4v HTTP/1.1" 206 1 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.107 Safari/537.36"

[14/May/2014:14:06:53 +0200] "GET /dev/test_test.m4v HTTP/1.1" 200 130680 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.10

[14/May/2014:14:07:15 +0200] "GET /dev/test_test.m4v HTTP/1.1" 206 1 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.107 Safari/537.36"

[14/May/2014:14:07:15 +0200] "GET /dev/test_test.m4v HTTP/1.1" 200 114684 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.10

You can see that theere is always one request with response size=1 and another one with mixed size, but >1. However, the result in the browser is always the same. => broken response, failed download.

For making sure, it's not related to http status/response 206, I add max_ranges 0; to the config and retry -> test2.

TEST2

[14/May/2014:14:11:36 +0200] "GET /dev/test_test.m4v HTTP/1.1" 200 152460 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.107 Safari/537.36"

[14/May/2014:14:11:38 +0200] "GET /dev/test_test.m4v HTTP/1.1" 200 142296 "-" "Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.107 Safari/537.36"

Same result in the browser. Always failed download/request and the size in access.log is always different, like random chunk or response size.

server status: iowait, RAM, cpu are almost idle. There is no high load or limitation.

Is this a known issue or bug or do You know how to fix this?

Apart from memory and cpu limitations, I'd like to keep in mind another number for the nginx and php-fpm setup.

If there is process.max set in /etc/php5/fpm/php-fpm.conf, the max number of php process ready for serving is limited.

However, nginx never knows about this limit and tries to find and get served by another php-fpm process.

How to limit nginx to request the same max at the same moment? (independent from cpu/ram which is discussed and explained in lots of other posts)

Using squid proxy, I'd like to (by)pass traffic from clients to a specific IP/domain.

No matter who is using the proxy (client IP or auth), the proxy should not be the man in the middle for any traffic from/to youtube.com.

Do proxies allow such a rule at server-side (instead of defining the bypass rule with each of the clients)?

E.g. if there was a bypass rule for the ip/domain whatismyip.com, all the clients should see their own IP there, although they have set up the proxy for their browser.

After trying to duplicate a nginx to another, I worry about nginx not starting.

nginx -t says that the config syntax is ok.

In nginx.conf it's told to log errors: error_log /var/log/nginx/error.log;

The error.log exists and it is re-created when starting, but always empty.

After /etc/init.d/nginx start or nginx , there is no notice or error in the shell and also not a line in the error.log. Anyways, ps -A|grep nginx tells me that there is nothing started.

Where to find an error message or how to find out what went wrong.

There seems to be no "verbose" mode for nginx :-(

Do I have an option to forbid a user group or a sftpuser to set chmod +x?

The user shall be allowed to upload media files, but no php scripts to run for example.

Can I set an /etc/fstab with options=auto to be skipped if mounting fails?

The manual says, it will break the boot process if an entry with options=auto fails.

Without auto, it does not automatically mount the device on boot.

So how to make the auto entry to be ignored on failures?

Since I separated my mailserver from the apache/webserver for the same domain, I worry about mails which are sent e.g. via php from the webserver to @same-domain.com. It's trying to deliver the message on localhost although the email server (target) is another machine :-(

How to fix the route to get the mails delivered?

Thanks

For a serverfarm with a tracking solution for several clients, I need a SSL certificate for multiple domains (3 different ones) and on each of the domains for all the subdomains.

I do only need it to get the tracking pixels without any browser warnings when embedded in https pages on client side. It is NOT used for security issues or so.

What might be the best product for my use-case?

Thanks

My webhoster seems to have some troubles with DDOS attacks and routing overload. This makes my IP to be not available sometimes and I'd like to add a failover IP for the domain.

However, 2 A-Records means that it's similar to a load balancer which is NOT what i am looking for. I do not want the fallback machine to be accessed any time - except in case of real troubles or downtime of first choice IP.

Is there a out-of-the box solution that allows me such a failover ip?

If not, if it requires a DNS re-configuration: Is there a recommended script or similar that can be run on the secondary machine (of course, this is with a different hoster). This has to check whether the first choice ip is down and must overwrite the A-record, right? This requires my domain hoster to provide an API for the DNS settings...

crm status is kind of confusing :-( Which of the 2 nodes is the currently selected one and which one is the alternative machine for failover scenario?

 crm status
 ============
 Last updated: Sun Jul 29 20:07:17 2012
 Stack: Heartbeat
 Current DC: testsrvm2.test-domain.biz (d5eb712f-d246-4363-a4e3-73fec90be179) - partition with quorum
 Version: 1.0.9-74392a28b7f31d7ddc86689598bd23114f58978b
 2 Nodes configured, 1 expected votes
 1 Resources configured.
 ============

 Online: [ testsrvm2.test-domain.biz testsrvm1.test-domain.biz ]

  Resource Group: IPFailover
      ip1        (ocf::heartbeat:IPaddr2):       Started testsrvm1.test-domain.biz
      ip1arp     (ocf::heartbeat:SendArp):       Started testsrvm1.test-domain.biz

Thx

Trying to re-sync a slave, I run into troubles with some of my tables which do not have a auto-increment primary index, but a primary unique ID (not auto-increment).

I run the re-sync with

mk-table-sync --execute --sync-to-master localhost --databases gman --table "list,of,my,problem,tables" --wait 60

This works for all tables with auto-inc primary ID, and for the others it says:

Can't make changes on the master because no unique index exists at /usr/bin/mk-table-sync line 7733. while doing db123.problem on localhost

How to fix this in another way? I do only want the slave to be synced with the proper data from master. If possible, i'd prefer doing this with mk-table-sync instead of LOCK, dump, insert the dump on the slave which is really old-fashioned and not very useful use-case.

Thanks for tips on maakit.

Does nginx have an option to ignore health checks or buckets from load balancer against nodes? There are thousands of entries like

93.190.2.176 - - [24/Jul/2012:08:56:59 +0200] "-" 400 0 "-" "-"

in my access.log which are useless to me.

Thanks for a tip

I found out that an imagemagick/convert application on my VPS is causing troubles due to high load on my box. Now i added renice with priority 10 to all of these convert-processes for throttling each minute.

However, I am not sure if it really slows down the process and relieves my CPU (only 1-core).

How to really make sure that specific processes do not increase the load significantly or attack the stable system?

I have >500MB free RAM, hardly any swap in use and a low iowait at any time. So I guess, it's really because of the cpu with only 1 core. Most of the time, this is enough and no waiting times. But in case of some conversions are running, it's failing.

On a VPS with 1 CPU core and 2GB RAM, I run a mysql+apache2 for a low traffic website. Sometimes the machine slows down or stops delivering through apache or mysql.

That's why I set up nagios which is sending me alerts like "Service Alert: localhost/Current Load is WARNING" after 5-10 days of running. Then I can login through SSH and check RAM with "free" which is still enough, 500MB+ available and only 60MB of swap in use.

Since the system slowed down again, I checked the syslog and found lots of these entries:

Jun 30 23:46:31 cl22 postfix/error[2190]: 46D8974323: to=, relay=none, delay=294806, delays=294803/3/0/0, dsn=4.4.3, status=deferred (delivery temporarily suspended: Host or domain name not found. Name service error for name=zombine.com type=MX: Host not found, try again) Jun 30 23:46:31 cl22 postfix/error[2193]: 49CB374123: to=, relay=none, delay=154189, delays=154185/3.1/0/0, dsn=4.4.3, status=deferred (delivery temporarily suspended: Host or domain name not found. Name service error for name=zombine.com type=MX: Host not found, try again) Jun 30 23:46:31 cl22 postfix/error[2153]: 4E2C874250: to=, relay=none, delay=433708, delays=433704/3.1/0/0, dsn=4.4.3, status=deferred (delivery temporarily suspended: Host or domain name not found. Name service error for name=zombine.com type=MX: Host not found, try again) Jun 30 23:46:31 cl22 postfix/error[2176]: 480D874180: to=, relay=none, delay=174308, delays=174304/3.1/0/0, dsn=4.4.3, status=deferred (delivery temporarily suspended: Host or domain name not found. Name service error for name=zombine.com type=MX: Host not found, try again)

How can I find out which process is consuming all the load? It's really lots of overload for a 1-core VPS: WARNING - load average: 3.06, 5.79, 3.42

mysql is OK, apache2 seems to be OK. postfix maybe not? anything else I did not identify yet?

Please let me know how to find out the bad process and temporarily renice or un-priorize postfix etc. to make sure that apache2 and mysql remain healthy. These 2 processes are important to me. The outgoing emails, too, because it's sending messages to clients.

After several nagios messages like "CRITICAL - load average: 135.12, 92.20, 44.09" and oom-killer, I checked the configuration again and again, with different settings, but no success.

The only work-around which seems to help on the ubuntu box, is this cleaner cronjob:

[ -x /usr/lib/php5/maxlifetime ] && [ -d /var/lib/php5 ] && find /var/lib/php5/ -depth -mindepth 1 -maxdepth 1 -type f -cmin +$(/usr/lib/php5/maxlifetime) ! -execdir fuser -s {} 2>/dev/null \; -delete

Reading the oom-killer info, tells me about apache2 to be the bad (parent) process. There is only few traffic on this vserver and hardly any peaks.

lscpu

Architecture: x86_64

CPU op-mode(s): 32-bit, 64-bit

Byte Order: Little Endian

CPU(s): 1

free -m

Mem: 2003 835 1168 0 18 381

-/+ buffers/cache: 435 1568

Swap: 1019 129 890

apache2.conf mpm

<IfModule mpm_prefork_module>
    StartServers          2
    MinSpareServers       2
    MaxSpareServers      10
    MaxClients          150
    MaxRequestsPerChild   10000
</IfModule>
<IfModule mpm_worker_module>
    StartServers          2
    MinSpareThreads      25
    MaxSpareThreads      75
    KeepAliveTimeout     15     
    ThreadLimit          64
    ThreadsPerChild      25
    MaxClients          100     
    MaxRequestsPerChild 10000   
</IfModule>
<IfModule mpm_event_module>
    StartServers          2
    MinSpareThreads      25
    MaxSpareThreads      75
    ThreadLimit          64
    ThreadsPerChild      25
    MaxClients          100     
    MaxRequestsPerChild 10000
</IfModule>

Isn't there something like a mysql_tuner.sh for apache or a useful example for a similar vps?

THX

In mongodb i work with this collection:

Size     55.93g 
Data Size    39.82g
Storage Size 41.08g 
Extents  53 
Indexes  4 
Index Size   9.64g

It takes few seconds of mongdb being up with this single collection and all 48GB RAM on the dedicated server are gone.

That's worse because there is also a mysqld+nginx/fcgi on this machine which should be allowed to use at least 24GB together. I.e. remaining 24GB, enough for the mongod!

However, it does not share in a fair way. Everybody says that the memory for mongod is managed by OS and releases unneccessary space for other processes if they demand RAM. On my machine it is not releasing RAM. What's wrong?

free

             total       used       free     shared    buffers     cached`
Mem:      49559136   49403908     155228          0      57284   47247564
-/+ buffers/cache:    2099060   47460076
Swap:      8008392        164    8008228

Since i faced issues with apache2 processes eating all my memory within short time, i found a workaround like this cronjob:

0,15,30,45 * * * * [ -x /usr/lib/php5/maxlifetime ] && [ -d /var/lib/php5 ] && find /var/lib/php5/ -depth -mindepth 1 -maxdepth 1 -type f -cmin +$(/usr/lib/php5/maxlifetime) ! -execdir fuser -s {} 2>/dev/null \; -delete

Before my RAM load is 2GB and after cronjob ran, it's 500MB load => 1.5GB free.

The cronjob is set to run on 0,15,30,45 every hour. Isn't there a proper solution which keeps the apache2 in check to avoid running out of memory within any 15min?

http://nopaste.info/a4de552b95.html

On a master-slave replication I have some shared tables which are written by master only.

In addition to the shared data, I have some "log tables" which are only relevant for the single machine, not to be shared and not to be backed-up.

If there is no mistake in the application on the server, it will not directly write into the slave machine - except the un-shared log tables. Isn't it possible to set the readonly flag for the shared tables explicitely? I found some similar questions, but no trustworthy solution yet :-(

Mistakes in the php {} part of vhost configuration files make php files to be "readable" with its source code like text files

Is there a global setting, e.g. in the default config to forbid accessing php files without fastcgi?