Building ELK on CentOS 7

scottalanmiller

Why lock out with .htaccess? There is no hint what is needed to log in here.

It's how Digital Ocean does it as well. Kibana doesn't have a built in authentication scheme that I know of. HTAccess is very simple for someone to just get started.

scottalanmiller

And simple to remove when you want to move to something else.

scottalanmiller

@JaredBusch said:

Line 109 needs commented out.

add this right after the yum install to fix the firewall.
yum -y install wget firewalld epel-release
systemctl enable firewalld
systemctl start firewalld
yum -y install nginx httpd-tools unzip
I would just remove line 109 it serves no purpose.

Edit: Some dumbass forgot to snapshot the image so he could repeat the install...

Thanks. That was formatting I had originally put in before scripting it.

scottalanmiller

@JaredBusch said:

Looks like maybe you forgot to start firewalld?

Fixed

JaredBusch

@scottalanmiller said:

@JaredBusch said:

@scottalanmiller so what do you setup your disk partitioning like in CentOS 7?

If I'm doing this for product, I do 20GB for the OS and 200GB+ on a second VHD for the data. I put it all under LVM and make a XFS filesystem on the secondary mount and mount it to data and make a symlink for the Elasticsearch database directory into there.

SO this mean you need to make one of your linux admin setrups on drive settings because that is not what CentOS does by dfault.

Dashrender

@JaredBusch said:

@scottalanmiller said:

@JaredBusch said:

@scottalanmiller so what do you setup your disk partitioning like in CentOS 7?

If I'm doing this for product, I do 20GB for the OS and 200GB+ on a second VHD for the data. I put it all under LVM and make a XFS filesystem on the secondary mount and mount it to data and make a symlink for the Elasticsearch database directory into there.

SO this mean you need to make one of your linux admin setrups on drive settings because that is not what CentOS does by dfault.

Would CentOS do what Scott does if you had two drives you provide CentOS to use? i.e. a 20 GB and a 200+ GB one? Would CentOS install the OS and everything fully on the 20, and then just mount the 200 on some point?

JaredBusch

@Dashrender said:

@JaredBusch said:

@scottalanmiller said:

@JaredBusch said:

@scottalanmiller so what do you setup your disk partitioning like in CentOS 7?

If I'm doing this for product, I do 20GB for the OS and 200GB+ on a second VHD for the data. I put it all under LVM and make a XFS filesystem on the secondary mount and mount it to data and make a symlink for the Elasticsearch database directory into there.

SO this mean you need to make one of your linux admin setrups on drive settings because that is not what CentOS does by dfault.

Would CentOS do what Scott does if you had two drives you provide CentOS to use? i.e. a 20 GB and a 200+ GB one? Would CentOS install the OS and everything fully on the 20, and then just mount the 200 on some point?

Than answer is not by default. It tries to make it's own magic.

You can see here I created a 20gb and a 200GB vhdx and told the install to handle it all for me.

0_1456324763628_upload-a8516602-b5e4-4c0e-9112-caabbb970b80

Guess what, you still end up with a 50GB and a 170GB partitions scheme

[root@elk ~]# df -h
Filesystem                   Size  Used Avail Use% Mounted on
/dev/mapper/centos_elk-root   50G  882M   50G   2% /
devtmpfs                     906M     0  906M   0% /dev
tmpfs                        916M     0  916M   0% /dev/shm
tmpfs                        916M  8.3M  907M   1% /run
tmpfs                        916M     0  916M   0% /sys/fs/cgroup
/dev/sda2                    494M   99M  395M  21% /boot
/dev/sda1                    200M  9.5M  191M   5% /boot/efi
/dev/mapper/centos_elk-home  168G   33M  168G   1% /home
tmpfs                        184M     0  184M   0% /run/user/0
[root@elk ~]#

JaredBusch

CentOS 7 has a thing for 50GB root mounts.

scottalanmiller

Yeah, the defaults suck a bit.

JaredBusch

@scottalanmiller Why are you using Oracle's Java SDK and not just java from the repo?

I had read in another write up on the install that it works fine even if it is not the "official" method.

scottalanmiller

@JaredBusch said:

@scottalanmiller Why are you using Oracle's Java SDK and not just java from the repo?

I had read in another write up on the install that it works fine even if it is not the "official" method.

Even if it "works", Elasticsearch tests against and only officially supports the Oracle one. Just because I can get the OpenJDK to work, I'd hate to have it be buggy or problematic for someone down the line because I used one that wasn't tested against.

JaredBusch

@scottalanmiller said:

@JaredBusch said:

@scottalanmiller Why are you using Oracle's Java SDK and not just java from the repo?

I had read in another write up on the install that it works fine even if it is not the "official" method.

Even if it "works", Elasticsearch tests against and only officially supports the Oracle one. Just because I can get the OpenJDK to work, I'd hate to have it be buggy or problematic for someone down the line because I used one that wasn't tested against.

I am not a fan of Oracle when it comes to Java

JaredBusch

@scottalanmiller

You hardcoded a DNS name in that script...

openssl req -subj '/CN=elk.lab.ntg.co/' -x509 -days 3650 -batch -nodes -newkey rsa:2048 -keyout private/logstash-forwarder.key -out certs/logstash-forwarder.crt

You also used a different port here for logstash than in the logstash forwarder example.

cat > /etc/logstash/conf.d/02-beats-input.conf <<EOF
input {
  beats {
    port => 5044
    ssl => true
    ssl_certificate => "/etc/pki/tls/certs/logstash-forwarder.crt"
    ssl_key => "/etc/pki/tls/private/logstash-forwarder.key"
  }
}
EOF

You used 5000 in that other post.

scottalanmiller

@JaredBusch said:

You used 5000 in that other post.

Which post was that? I bet that one was a typo, 5044 is the standard port.

JaredBusch

@scottalanmiller said:

@JaredBusch said:

You used 5000 in that other post.

Which post was that? I bet that one was a typo, 5044 is the standard port.

It was this post

I would also like to note that the certificate your script in this post creates will not work with your logstash forwarder instructions by IP address.

2016/02/24 13:21:48.989719 Failed to tls handshake with 10.201.1.16 x509: cannot validate certificate for 10.201.1.16 because it doesn't contain any IP SANs

JaredBusch

As soon as I corrected all of those issues, I got this.

2016/02/24 13:42:26.611248 Connecting to [10.201.1.16]:5044 (elk.domain.local)
2016/02/24 13:42:28.167827 Connected to 10.201.1.16
2016/02/24 13:42:32.038421 Registrar: processing 1024 events
2016/02/24 13:42:33.923706 Registrar: processing 1024 events
2016/02/24 13:42:35.424984 Registrar: processing 891 events
2016/02/24 13:45:55.815543 Registrar: processing 3 events
2016/02/24 13:46:03.305215 Registrar: processing 1 events

scottalanmiller

@JaredBusch said:

@scottalanmiller said:

@JaredBusch said:

You used 5000 in that other post.

Which post was that? I bet that one was a typo, 5044 is the standard port.

It was this post

I would also like to note that the certificate your script in this post creates will not work with your logstash forwarder instructions by IP address.
2016/02/24 13:21:48.989719 Failed to tls handshake with 10.201.1.16 x509: cannot validate certificate for 10.201.1.16 because it doesn't contain any IP SANs

Oh!! That post is for a different era using different tools. Can't use that with this. We are on the "beat" system now. A lot has changed, that's why I was doing the new write up as there was a lot new in the last few weeks that prompted new documentation.

JaredBusch

@scottalanmiller said:

@JaredBusch said:
@scottalanmiller said:

@JaredBusch said:

You used 5000 in that other post.

Which post was that? I bet that one was a typo, 5044 is the standard port.

It was this post

I would also like to note that the certificate your script in this post creates will not work with your logstash forwarder instructions by IP address.
2016/02/24 13:21:48.989719 Failed to tls handshake with 10.201.1.16 x509: cannot validate certificate for 10.201.1.16 because it doesn't contain any IP SANs
Oh!! That post is for a different era using different tools. Can't use that with this. We are on the "beat" system now. A lot has changed, that's why I was doing the new write up as there was a lot new in the last few weeks that prompted new documentation.

Well, i never did manage to get something working with beat. Your different era was only 8 months ago.

scottalanmiller

@JaredBusch said:

Well, i never did manage to get something working with beat. Your different era was only 8 months ago.

Well that's how paradigm changes work. ELK introduced a new scheme less than eight months ago. Whether it was a decade, eight months or two days ago, once they've changed their underlying architecture it's changed. It's not a gradual evolution over the years, it's one day they used one system and with the next release they used another. So at some point people installing "current" one minute had one thing and the next minute would have had the other. Has to change at some point in time.

scottalanmiller

Updated with information on generating the logstash certificate.