Pavel Tankov's questions

On CentOS7: I need to upgrade MariaDB from 5.5 to 10.4

I followed this guide: https://itcloudnet.blogspot.com/2019/10/how-to-upgrade-mariadb-55-to-mariadb.html but instructions on other sites are similar.

When I do

yum remove mariadb-server mariadb-devel mariadb-libs mariadb

I get

Removing:
 mariadb
 mariadb-devel
 mariadb-libs
 mariadb-server
Removing for dependencies:
 perl-DBD-MySQL
 php-mysql
 postfix

Then I proceed with the instructions and do

yum install MariaDB-server MariaDB-client

and the output is

Installing:
 MariaDB-client
 MariaDB-server
Installing for dependencies:
 MariaDB-common
 MariaDB-compat
 boost-program-options
 galera-4

So, no php-mysql here. And the biggest mystery of all is that the PHP web applications using MariaDB still work!

How come? So, after all, do I need the package php-mysql or not?

Some clarifications:

1. I use the standard Apache that comes with CentOS7 httpd-2.4.6.
2. We have two legacy PHP applications that use the legacy PHP version which comes standard with CentOS7, namely php-5.4.
3. We have one newer PHP application (Mediawiki) that needs newer PHP. So, I have installed php71 alongside with php-5.4, using Remi repository.
4. I have switched Apache from mod_php to php-fpm.
5. I have followed some online instructions how to make both legacy php-fpm-4.5 and the newer php71-php-fpm-7.1 (from Remi) work simultaneously on Apache - to serve both the two legacy and one newer web applications.

But my question still remains: Why my legacy PHP Web applications still work fine without the package php-mysql?

I have 3 hard disks 1T each. I want to wipe them out completely. I decided yo use dd like so:

dd if=/dev/zero of=/dev/sdb bs=1M
dd if=/dev/zero of=/dev/sdc bs=1M
dd if=/dev/zero of=/dev/sdd bs=1M

Since it takes a while for the commands to complete, can I assume that it will be faster if I open 3 different consoles and execute the commands simultaneously?

Will it be 3 times faster compared to executing them sequentially?

I am planning to setup a CentOS7 server with software RAID. Since the CentOS7 installer isn't particularly good at preparing the disks, I wonder if it would be OK to boot up a LinuxMint 18 live, do the partitioning and RAID assignment and then reboot in the CentOS7 installer and do the installation, skipping the partitioning part.

What worries me is that they seem to be packed with different versions of software for manipulating disks.

For example:

CentOS 7.4:

gdisk 0.8.6

mdadm 4.0-5

LinuxMint 18.3

gdisk 1.0.1

mdadm 3.3-2

1. So, if I create the raid array on Mint live with an older version of mdadm, should I expect problems when CentOS tries to assemble that array with a newer version of mdadm?
2. Will I be missing on features when I create the array with an older version of mdadm, like different default values, etc.?
3. How about gdisk, but in the opposite direction?

I plan to do a server with 2 SSDs in RAID1 (for OS and boot) and 3 HDDs in RAID5 (for data). All Linux (CentOS7) with mdadm.

We have in the office 2 leftover SSDs, but they are from different manufacturers and with different capacities (120G and 250G).

I have been reading about RAID1 on SSDs lately and my worries are:

1. When I create the raid, mdadm will insist on syncing it.

   1. Both SSDs have been used before. What exactly will mdadm sync when the cells of the two SSDs contain different data anyway?
   2. Why is that syncing even necessary?
   3. Does it shorten SSD's life (writing on all cells)?

2. The fstrim command will be issued by a weekly cron job on /, which will be the mount point of /dev/md0, which will be the RAID1. Recent versions of mdadm will pass through the trim to the individual hard disks themselves. However, I have read that different SSDs may have different strategies when trimming. For example, one SSD may immediately fill the cells with zeroes, while the other may not. That would mean that the two disks are now out of sync. Now, mdadm has a cron job that, by default, runs periodically and checks the consistency of RAID1 arrays. Will that check report failure because of different trim behaviour?

I am going to put a used SSD into a server to boot from it and hold the OS itself, while making a RAID1 from two other HDDs.

Since the SSD has been used, do I need to trim it beforehand and how? Or do I just shove it in the server and use it right away?

I have built software RAID once in the past (CentOS 7 and mdadm) and I know I was able to configure it to send an email when a disk fails.

Now I need to deal with an old server (about 10 years old) and it has hardware RAID. I configured it through its BIOS-like menus, but I wonder how it's going to tell me if there is a problem? There are no configuration options about mail or anything.

Many web services require a database and quite often that is MySQL. Typically, the installation instruction of the software tells you to create a database and a user with permission for that database.

Is it OK to name the user and the DB the same?

For example, I am installing Openfire right now and I created a MySQL database "openfire" and MySQL user "openfire". Should I expect any problems? What is the convention?

We have quite an old installation of SUSE LINUX 10.1 (i586) in the office.

The problem shortly: I can successfully ssh to it from machines in the same LAN (192.168.1.0) but not from machines in the other LAN (10.23.0.0).

The SuSE has SSH server openssh-4.2p1-18.12. I have ruled out the firewall and hosts.allow and hosts.deny files.

When my ssh login attempt fails, here is what the logs say:

on the client:

$ ssh -vvv 192.168.1.5
OpenSSH_5.3p1, OpenSSL 1.0.1e-fips 11 Feb 2013
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: Applying options for *
debug2: ssh_connect: needpriv 0
debug1: Connecting to 192.168.1.5 [192.168.1.5] port 22.
debug1: Connection established.
debug1: identity file /home/nbuild/.ssh/identity type -1
debug1: identity file /home/nbuild/.ssh/identity-cert type -1
debug1: identity file /home/nbuild/.ssh/id_rsa type -1
debug1: identity file /home/nbuild/.ssh/id_rsa-cert type -1
debug1: identity file /home/nbuild/.ssh/id_dsa type -1
debug1: identity file /home/nbuild/.ssh/id_dsa-cert type -1

on the server:

Aug 21 16:34:25 serverhost sshd[20736]: debug3: fd 4 is not O_NONBLOCK
Aug 21 16:34:25 serverhost sshd[20736]: debug1: Forked child 20739.
Aug 21 16:34:25 serverhost sshd[20736]: debug3: send_rexec_state: entering fd = 7 config len 403
Aug 21 16:34:25 serverhost sshd[20736]: debug3: ssh_msg_send: type 0
Aug 21 16:34:25 serverhost sshd[20736]: debug3: send_rexec_state: done
Aug 21 16:34:25 serverhost sshd[20739]: debug1: rexec start in 4 out 4 newsock 4 pipe 6 sock 7
Aug 21 16:34:25 serverhost sshd[20739]: debug1: inetd sockets after dupping: 3, 3
Aug 21 16:34:25 serverhost sshd[20739]: debug3: Normalising mapped IPv4 in IPv6 address
Aug 21 16:34:25 serverhost sshd[20739]: Connection from 10.23.1.11 port 44340

The above log on the server is when I enable DEBUG3 log level. However, with the default log level (INFO), the only thing the server logs is this:

Aug 21 16:38:32 serverhost sshd[20749]: Did not receive identification string from 10.23.1.11

Any hints? I feel I've tried everything already.

Update: The machines that cannot ssh are in another VLAN if that matters. I have tried CentOS 6.5 and Ubuntu.

I have installed OpenNebula following the quickstart guide OpenNebula on CentOS 6 and KVM.

I am using all local hard disks. I have one huge software RAID10 array composed of 4 partitions, no LVM whatsoever. The raid device is /dev/md2 and is mounted on /var/lib/one/datastores.

Here is the disk space as reported by CentOS:

$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/md2        7.1T  177M  6.8T   1% /var/lib/one/datastores

and here is the disk usage as reported by CentOS:

$ du -shc /var/lib/one/datastores
20K     /var/lib/one/datastores
20K     total

so, the folder seems pretty much empty (although I still don't know why df -h says that 177M are used, but anyway).

and here is the contents of the folder:

$ ls -la /var/lib/one/datastores
total 24
drwxr-xr-x 6 oneadmin oneadmin 4096 Jul 31 18:19 .
drwx------ 8 oneadmin oneadmin 4096 Jul 31 18:33 ..
drwxrwxr-x 2 oneadmin oneadmin 4096 Jul 31 18:20 0
drwxr-xr-x 2 oneadmin oneadmin 4096 Jul 31 18:20 1
drwxr-xr-x 2 oneadmin oneadmin 4096 Jul 31 18:04 2
drwxr-xr-x 2 oneadmin oneadmin 4096 Jul 31 18:13 .isofiles

$ find /var/lib/one/datastores
/var/lib/one/datastores
/var/lib/one/datastores/2
/var/lib/one/datastores/.isofiles
/var/lib/one/datastores/1
/var/lib/one/datastores/0

The problem is that OpenNebula says my datastores are using 367.8GB when they should be empty. See the screen shot.

Update: I have tried stopping OpenNebula, deleting the database /var/lib/one/one.db and starting OpenNebula again with no positive result.

OpenNebula uses the notions of "Images" and "Templates". From what I've been able to understand so far, an image is first downloaded from the marketplace, then a template is created from it, and finally a virtual machine is created from the template.

This seems a bit superfluous to me. Why have both images and templates? What's the difference? Why not a two-step process like: get an image (or template?) and then create a VM from it?

I am making a huge RAID10 device (8 TB) for storage on CentOS. I have created the 4 partitions on my 4 physical hard disks and have created the raid device /dev/md2 out of those 4 partitions.

Now it's time to put a filesystem on it. I am wondering whether I should do just:

mkfs.ext4 /dev/md2

which totally works (I tried it), or I should first create a single partition on that device (with fdisk, gdisk or parted?), that takes up all the space, and then do the mkfs stuff on that partition.

Googling the Web, I found another person asking the very same question here: http://ubuntuforums.org/showthread.php?t=2174067 but, although a responder said:

Partitions go on the array members.
Filesystem goes on the array.

they didn't give him a definitive, non-ambiguous answer or an explanation when one approach should be preferred over the other.