Home / user-269679

gxx's questions

Martin Hope
gxx
Asked: 2017-10-01 01:27:42 +0800 CST
  • 0

I've got a server running open source Nginx 1.10.3 with a pretty simple config: Serving static files, in case a file doesn't exist, the request is send to a backend.

I need to limit the active connections to this backend. Quite easy, isn't it, using upstream server max_conns, which

limits the maximum number of simultaneous active connections to the proxied server.

Besides, this was only introduced in 1.11.5, so it's not (yet) available for me.

Any clever ideas / workarounds out there dealing with this situation?

[This server is currently running Debian jessie with nginx out of jessie-backports. I could upgrade to stretch, and use stretch-backports. But, before doing so, and because I'm quite happy with this system, I would like to check if there is any other way around. You know: Never change a running system...]

nginx
Martin Hope
gxx
Asked: 2016-09-03 05:12:12 +0800 CST
  • 3

If one thinks of a better title for this question: Please change it accordingly.

Environment:

  • Debian jessie 8.5
  • puppet master: 3.8.5
  • puppet agent: 3.7.2
  • apt: 1.0.9.8.3

Problem:

A specific package is installed already on the node, let's take needrestart for example:

# dpkg -l | grep needrestart
ii  needrestart                     1.2-8+deb8u1

Using the following code:

package{[
  'needrestart',
]:
  ensure          => installed, # latest won't work, either
  install_options => ['-t', 'jessie-backports']
}

I would expect that needrestart is reinstalled / upgraded to the version out of jessie-backports. However, this doesn't happen, there is no reinstall / upgrade happening. Inside the (debug) logs there's nothing worth to be shown regarding this.

Solution:

In case this is not intended: Anyone knows a clever way around this?

debian puppet apt
Martin Hope
gxx
Asked: 2015-02-11 07:33:20 +0800 CST
  • 3

The machine in question is equipped with 4x ssd, and running Debian wheezy 7.8, linux kernel 3.2.0-4, mdadm v3.2.5. Currently I'm using a RAID10 consisting of all these ssd. I've to increase the storage capacity, ssd are to expensive for this, and so thought about getting rid of two ssd, instead putting two hdd inside, and splitting the RAID10 into two RAID1, one made out of the ssd, the other out of the hdd.

I've searched the Internet a lot these past days, but not really found anything useful regarding this. Could someone tell me how to convert a RAID10 into one (or two) RAID1?

I'm quite unsure if this is possible at all, because, for example, Neil Brown, one of the mdadm developers, wrote 2009 [1] (which is a while back, sure) about reshaping:

RAID0 and RAID10

These arrays cannot be reshaped at all at present.

Could someone tell me if this is possible nowadays and if so, how to do this?

Thanks in advance!

[1] http://neil.brown.name/blog/20090817000931#4

Edit 1: At [2], someone was asking if converting raid10 to raid0 is possible, getting three answers, one of them being "Yes.", one being "No." - quite confusing.

[2] Can I convert a raid10 array to raid0?

Edit 2: Using mdadm /dev/md0 --grow --level=1 (in a virtual test machine) gives mdadm: RAID10 can only be changed to RAID0. So it seems the way to go is converting to RAID0, and afterwards to RAID1; but don't like this idea really, because it's quite risky... Anyone knows of a better, safer, process?

Edit 3: It's not possible to convert a RAID0 to RAID1, trying to do so gives mdadm: Impossibly level change request for RAID1. So it seems one would then have to create a new array with level 1, made up of the two devices currently serving the RAID0.

linux