{{>toc}}

h1. Management Cluster Ceph

h2. Liens

* [[Openstack Management TTNN]]

* [[Openstack Setup VM pas dans openstack]]

* [[Openstack Installation nouvelle node du cluster]]

* [[Openstack Installation TTNN]]

* "Openstack tools for ttnn":/projects/git-tetaneutral-net/repository/openstack-tools

h2. Ajout d'un OSD classique

<pre>

$ ceph-disk prepare --zap-disk --cluster-uuid 1fe74663-8dfa-486c-bb80-3bd94c90c967 --fs-type=ext4 /dev/sdX

$ tune2fs -c 0 -i 0 -m 0 /dev/sdX1

$ smartctl --smart=on /dev/sdX  # Pour le monitoring.

</pre>

Récuperer l'id avec (c'est celui tout en bas pas accroché à l'arbre):

<pre>

ceph osd tree

</pre>

*DEBUT WORKAROUND BUG PREPARE*

Dans le cas ou l'osd est DOWN après le prepare c'est surement ce bug

ID est le premier numero libre d'osd en partant de zero (en bas du ceph osd tree)

<pre>

mkdir /var/lib/ceph/osd/ceph-<ID>

chown ceph:ceph /var/lib/ceph/osd/ceph-<ID>

ceph-disk activate /dev/sd<X>1

systemctl status ceph-osd@<ID>

</pre>

*FIN WORKAROUND BUG PREPARE*

Pour un HDD: 

<pre>

$ ceph osd crush add osd.<ID> 0 root=default host=<host>

</pre>

Pour un SSD:

<pre>

$ ceph osd crush add osd.<ID> 0 root=ssd host=<host>-ssd

</pre>

Ensuite, autoriser Ceph à mettre des data dessus:

<pre>

$ /root/tools/ceph-reweight-osds.sh osd.<ID>

</pre>

h3. helper

<pre>

add_osd(){

  dev="$1"

  type="$2"

  host=$(hostname -s)

  [ "$type" == "ssd" ] && host="${host}-ssd"

  next=-1 ; for i in $(ceph osd ls); do next=$((next+1)) ; [ $(($i - $next)) -gt 0 ] && break; done ;

  mkdir /var/lib/ceph/osd/ceph-$next

  chown ceph:ceph /var/lib/ceph/osd/ceph-$next

  ceph-disk prepare --zap-disk --cluster-uuid 1fe74663-8dfa-486c-bb80-3bd94c90c967 --fs-type=ext4 $dev

  tune2fs -c 0 -i 0 -m 0 ${dev}1

  smartctl --smart=on $dev

  systemctl start ceph-osd@$next

  systemctl status ceph-osd@$next

  ceph osd crush add osd.$next 0 root=${type} host=${host}

}

</pre>

h2. Vider un OSD:

<pre>

vider_osd(){

  name="$1"

  ceph osd out ${name}

  ceph osd crush reweight ${name} 0

  ceph osd reweight ${name} 0

}

</pre>

h2. Suppression d'un OSD:

<pre>

remove_osd(){

 name="$1" 

 ceph osd out ${name}

 systemctl stop ceph-osd@${name#osd.}

 ceph osd crush remove ${name}

 ceph auth del ${name}

 ceph osd rm ${name}

 ceph osd tree

}

</pre>

h2. Arrêter les IO de recovery:

<pre>

ceph osd set nobackfill 

ceph osd set norebalance

ceph osd set norecover

</pre>

h2. Procédure d'upgrade

+

_*/!\Lire la release note (contient très très souvent des trucs à faire en plus) /!\*_+

h4. Upgrade des MONs:

Mettre le flags noout:

<pre>ceph osd set noout</pre>

Sur chaque MONs (g1/g2/g3)

<pre>

apt-get upgrade -y

systemctl restart ceph-mon@g*

ceph -s

</pre>

Note que seulement le node 'leader/master' va provoquer une micro/nano coupure, souvent c'est même invisible.

h4. Upgrade des OSDs:

Pour chaque machine

<pre>

apt-get upgrade -y

systemctl restart ceph-osd@*

</pre>

Puis attendre que le recovery termine avant de faire la suivante.

Une fois toutes les OSDs upgrader et relancer, faire:

<pre>ceph osd unset noout</pre>

h2. Remplacement à froid d'un tier cache:

upstream  doc: http://docs.ceph.com/docs/master/rados/operations/cache-tiering/

<pre>

ceph osd tier cache-mode ec8p2c forward

rados -p ec8p2c cache-flush-evict-all

ceph osd tier remove-overlay ec8p2

ceph osd tier remove ec8p2 ec8p2c

rados rmpool ec8p2c ec8p2c  --yes-i-really-really-mean-ita

ceph osd pool create ec8p2c 128 128 replicated

ceph osd tier add ec8p2 ec8p2c

ceph osd tier cache-mode ec8p2c writeback

ceph osd tier set-overlay ec8p2 ec8p2c

ceph osd pool set ec8p2c size 3

ceph osd pool set ec8p2c min_size 2

ceph osd pool set ec8p2c hit_set_type bloom

ceph osd pool set ec8p2c hit_set_count 1

ceph osd pool set ec8p2c hit_set_period 3600

ceph osd pool set ec8p2c target_max_bytes 200000000000

ceph osd pool set ec8p2c target_max_objects 10000000

ceph osd pool set ec8p2c cache_target_dirty_ratio 0.4

ceph osd pool set ec8p2c cache_target_full_ratio 0.8

</pre>

h2. Ajout d'un OSD qui partage le SSD avec l'OS (OBSOLETE PLUS COMPATIBLE AVEC LES FUTURES VERSION DE CEPH)

En général avec ceph, on donne un disque, ceph créé 2 partitions une pour le journal de l'OSD, l'autre pour les datas

mais pour le SSD de tetaneutral qui a aussi l'OS, voici la méthode

Création manuelle de la partition de data ceph /dev/sda2 ici

Debian (MBR format):

<pre>

apt-get install partprobe

fdisk /dev/sda

n

p

<enter>

<enter>

<enter>

<enter>

w

$ partprobe

</pre>

Ubuntu (GPT format):

<pre>

# parted /dev/sdb

GNU Parted 2.3

Using /dev/sdb

Welcome to GNU Parted! Type 'help' to view a list of commands.

(parted) print                                                            

Model: ATA SAMSUNG MZ7KM480 (scsi)

Disk /dev/sdb: 480GB

Sector size (logical/physical): 512B/512B

Partition Table: msdos

Number  Start   End     Size    Type     File system     Flags

 1      1049kB  20.0GB  20.0GB  primary  ext4            boot

 2      20.0GB  36.0GB  16.0GB  primary  linux-swap(v1)

(parted) mkpart                                                           

Partition type?  primary/extended?                                        

Partition type?  primary/extended? primary

File system type?  [ext2]? xfs                                            

Start?                                                                    

Start? 36.0GB                                                             

End? 100%                                                                 

(parted) print                                                            

Model: ATA SAMSUNG MZ7KM480 (scsi)

Disk /dev/sdb: 480GB

Sector size (logical/physical): 512B/512B

Partition Table: msdos

Number  Start   End     Size    Type     File system     Flags

 1      1049kB  20.0GB  20.0GB  primary  ext4            boot

 2      20.0GB  36.0GB  16.0GB  primary  linux-swap(v1)

 3      36.0GB  480GB   444GB   primary

(parted) quit                                                             

Information: You may need to update /etc/fstab.

</pre>

On prepare le disk comme normalement

<pre>

ceph-disk prepare --fs-type=ext4 --cluster-uuid 1fe74663-8dfa-486c-bb80-3bd94c90c967 /dev/sda2

ceph-disk activate /dev/sda2

ceph osd crush add osd.<ID> 0 root=ssd host=g3-ssd

</pre>

Ensuite, autoriser Ceph à mettre des data dessus:

<pre>

$ /root/tools/ceph-reweight-osds.sh osd.<ID>

</pre>

h2. inconsistent pg

* Analyse d'une erreur de coherence detectee par ceph

** https://lists.tetaneutral.net/pipermail/technique/2017-August/002859.html

<pre>

root@g1:~# ceph health detail

HEALTH_ERR 1 pgs inconsistent; 2 scrub errors

pg 58.22d is active+clean+inconsistent, acting [9,47,37]

2 scrub errors

root@g1:~# rados list-inconsistent-obj 58.22d  --format=json-pretty

{

    "epoch": 269000,

    "inconsistents": [

        {

            "object": {

                "name": "rbd_data.11f20f75aac8266.00000000000f79f9",

                "nspace": "",

                "locator": "",

                "snap": "head",

                "version": 9894452

            },

            "errors": [

                "data_digest_mismatch"

            ],

            "union_shard_errors": [

                "data_digest_mismatch_oi"

            ],

            "selected_object_info": 

"58:b453643a:::rbd_data.11f20f75aac8266.00000000000f79f9:head(261163'9281748 osd.9.0:6221608 dirty|data_digest|omap_digest s 4194304 uv 9894452 dd 2193d055 od ffffffff alloc_hint [0 0])",

            "shards": [

                {

                    "osd": 9,

                    "errors": [],

                    "size": 4194304,

                    "omap_digest": "0xffffffff",

                    "data_digest": "0x2193d055"

                },

                {

                    "osd": 37,

                    "errors": [

                        "data_digest_mismatch_oi"

                    ],

                    "size": 4194304,

                    "omap_digest": "0xffffffff",

                    "data_digest": "0x05891fb4"

                },

                {

                    "osd": 47,

                    "errors": [],

                    "size": 4194304,

                    "omap_digest": "0xffffffff",

                    "data_digest": "0x2193d055"

                }

            ]

        }

    ]

}

root@g1:~# ceph osd map disks rbd_data.11f20f75aac8266.00000000000f79f9

osdmap e269110 pool 'disks' (58) object 'rbd_data.11f20f75aac8266.00000000000f79f9' -> pg 58.5c26ca2d (58.22d) -> up ([9,47,37], p9) acting ([9,47,37], p9)

root@g8:/var/lib/ceph/osd/ceph-9/current/58.22d_head# find . -name '*11f20f75aac8266.00000000000f79f9*'

./DIR_D/DIR_2/DIR_A/DIR_C/rbd\udata.11f20f75aac8266.00000000000f79f9__head_5C26CA2D__3a

root@g10:/var/lib/ceph/osd/ceph-37/current/58.22d_head# find . -name '*11f20f75aac8266.00000000000f79f9*'

./DIR_D/DIR_2/DIR_A/DIR_C/rbd\udata.11f20f75aac8266.00000000000f79f9__head_5C26CA2D__3a

$ scp g8:/var/lib/ceph/osd/ceph-9/current/58.22d_head/DIR_D/DIR_2/DIR_A/DIR_C/rbd*data.11f20f75aac8266.00000000000f79f9__head_5C26CA2D__3a g8data

$ scp g10:/var/lib/ceph/osd/ceph-37/current/58.22d_head/DIR_D/DIR_2/DIR_A/DIR_C/rbd*data.11f20f75aac8266.00000000000f79f9__head_5C26CA2D__3a g10data

$ md5sum *

bd85c0ef1f30829ce07e5f9152ac2d2f  g10data

4297d0bc373e6603e0ad842702e0ecaa  g8data

$ $ diff -u <(od -x g10data) <(od -x g8data)

--- /dev/fd/63	2017-08-13 10:43:52.837097740 +0200

+++ /dev/fd/62	2017-08-13 10:43:52.833097808 +0200

@@ -2617,7 +2617,7 @@

 0121600 439b 14f4 bb4c 5f14 6ff7 4393 9ff8 a9a9

 0121620 29a8 56a4 1133 b6a8 2206 4821 2f42 4b2c

 0121640 3d86 41a2 785f 9785 8b48 4243 e7b9 f0aa

-0121660 29b6 be0c 0455 bf97 1c0d 49e5 75dd e1ed

+0121660 29a6 be0c 0455 bf97 1c0d 49e5 75dd e1ed

 0121700 2519 d6ac 1047 1111 0344 38be 27a1 db07

 0121720 dff6 c002 75d8 4396 6154 eba9 3abd 5d20

 0121740 8ae4 e63a 298b d754 0208 9705 1bb8 3685

</pre>

Donc un seul bit flip 29b6 vs 29a6  

<pre>

>>> bin(0xa)

'0b1010'

>>> bin(0xb)

'0b1011'

</pre>

* http://cephnotes.ksperis.com/blog/2013/08/20/ceph-osd-where-is-my-data

* https://superuser.com/questions/969889/what-is-the-granularity-of-a-hard-disk-ure-unrecoverable-read-error