ʻO Docker Swarm, Kubernetes, a me Mesos ka papa hana hoʻoheheʻe ipu kaulana loa. Ma kāna kamaʻilio ʻana, hoʻohālikelike ʻo Arun Gupta i nā ʻano o Docker, Swarm, a me Kubernetes:
- Hoʻomohala kūloko.
- Nā hana hoʻolaha.
- Nā noi pahu nui.
- ʻIke lawelawe.
- Hoʻonui i ka lawelawe.
- Nā hana holo hoʻokahi.
- Hoʻohui me Maven.
- "Rolling" hou.
- Ke hana ʻana i kahi pūʻulu waihona Couchbase.
ʻO ka hopena, e loaʻa iā ʻoe kahi ʻike maopopo i ka mea a kēlā me kēia mea hana hoʻokani pila e hāʻawi ai a aʻo i ka hoʻohana pono ʻana i kēia mau kahua.
ʻO Arun Gupta ka mea ʻenehana nui no nā huahana open-source ma Amazon Web Services, nāna i hoʻomohala i nā kaiāulu o ka Sun, Oracle, Red Hat a me Couchbase no nā makahiki he 10. Loaʻa i ka ʻike nui e hana ana ma ke alakaʻi ʻana i nā hui cross-functional e hoʻomohala a hoʻokō i ka hoʻolālā no nā hoʻolaha kūʻai aku a me nā papahana. Ua alakaʻi ʻo ia i nā hui o Sun engineers, ʻo ia kekahi o nā mea hoʻokumu o ka hui Java EE a me ka mea nāna i hana i ka lālā US o Devoxx4Kids. ʻO Arun Gupta ka mea kākau ma mua o 2 tausani mau pou ma IT blogs a ua hāʻawi i nā kamaʻilio ma mua o 40 mau ʻāina.
Loaʻa i ka laina 55 kahi COUCHBASE_URI e kuhikuhi ana i kēia lawelawe waihona, i hana ʻia me ka hoʻohana ʻana i ka faila hoʻonohonoho Kubernetes. Inā ʻoe e nānā i ka laina 2, hiki iā ʻoe ke ʻike i ke ʻano: Service ka lawelawe aʻu e hana nei i kapa ʻia ʻo couchbase-service, a ua helu ʻia ka inoa like ma ka laina 4. Aia ma lalo kekahi mau awa.
ʻO nā laina kī ʻo 6 a me 7. Ma ka lawelawe ʻana wau, "E, eia nā lepili aʻu e ʻimi nei!", A ʻo kēia mau lepili he mea ʻē aʻe ma mua o nā inoa paʻa like ʻole, a me nā laina 7 i koʻu couchbase-rs-pod. palapala noi. Eia nā awa e hiki ai ke komo i kēia mau lepili like.
Ma ka laina 19 hana wau i kahi ReplicaSet hou, laina 31 ka inoa o ke kiʻi, a kuhikuhi nā laina 24-27 i ka metadata pili me kaʻu pod. ʻO kēia ka mea e ʻimi nei ka lawelawe a me ka mea e pono ai ka pilina. Ma ka hope o ka faila aia kekahi ʻano pilina ma waena o nā laina 55-56 a me 4, e ʻōlelo ana: "hoʻohana i kēia lawelawe!"
No laila, hoʻomaka wau i kaʻu lawelawe inā loaʻa kahi replica set, a no ka mea aia kēlā me kēia set replica i kona awa ponoʻī me ka lepili pili, ua hoʻokomo ʻia i ka lawelawe. Mai ka manaʻo o ka mea hoʻomohala, kāhea wale ʻoe i ka lawelawe, a laila hoʻohana i ka hoʻonohonoho o nā replicas āu e pono ai.
ʻO ka hopena, loaʻa iaʻu kahi pod WildFly e kamaʻilio me ka backend waihona ma o Couchbase Service. Hiki iaʻu ke hoʻohana i ka frontend me kekahi mau WildFly pods, e kamaʻilio pū me ka couchbase backend ma o ka lawelawe couchbase.
Ma hope e nānā mākou pehea e kamaʻilio ai kahi lawelawe ma waho o ka cluster ma o kāna IP address me nā mea i loaʻa i loko o ka pūʻulu a loaʻa kahi IP IP kūloko.
No laila, maikaʻi nā ipu mokuʻāina, akā pehea ka maikaʻi o ka hoʻohana ʻana i nā ipu stateful? E nānā kāua i nā hoʻonohonoho ʻōnaehana no nā pahu mokuʻāina, a i ʻole nā mea hoʻomau. Ma Docker, aia he 4 mau ala ʻokoʻa i ka hoʻonohonoho mālama ʻikepili āu e hoʻolohe ai. ʻO ka mea mua ʻo Implicit Per-Container, ʻo ia ka mea i ka wā e hoʻohana ai i ka couchbase, MySQL a i ʻole MyDB satateful ipu, hoʻomaka lākou āpau me ka Sandbox paʻamau. ʻO ia hoʻi, mālama ʻia nā mea a pau i mālama ʻia i loko o ka waihona. Inā nalowale ka ipu, nalowale ka ʻikepili me ia.
ʻO ka lua ka Explicit Per-Container, ke hana ʻoe i kahi waihona kikoʻī me ka docker volume e hana i ke kauoha a mālama i ka ʻikepili i loko. Hoʻopili ʻia ke kolu o ka Per-Host me ka hoʻopaʻa ʻana i ka palapala ʻāina, i ka manawa i hoʻopaʻa ʻia nā mea a pau i mālama ʻia i loko o ka pahu i ka manawa hoʻokahi. Inā hāʻule ka pahu, e hoʻomau ka ʻikepili ma ka host. ʻO ka mea hope ka hoʻohana ʻana i kekahi mau pūʻali Multi-Host, i ʻōlelo ʻia i ka pae hana o nā ʻano hoʻonā like ʻole. E ʻōlelo kākou e holo ana kāu mau pahu me kāu mau noi ma ka mea hoʻokipa, akā makemake ʻoe e mālama i kāu ʻikepili ma kahi o ka Pūnaewele, a no kēia mea e hoʻohana ai ʻoe i ka palapala 'akomi no nā ʻōnaehana māhele.
Ke hoʻohana nei kēlā me kēia ʻano hana i kahi wahi mālama pono. Hoʻopaʻa ʻia nā ʻikepili Per-Container implicit a Explicit ma ka host ma /var/lib/docker/volumes. I ka hoʻohana ʻana i ke ʻano Per-Host, kau ʻia ka waihona i loko o ka pahu, a kau ʻia ka pahu ma luna o ka mea hoʻokipa. No nā multihosts, hiki ke hoʻohana ʻia nā hoʻonā e like me Ceph, ClusterFS, NFS, etc.
Inā hāʻule ka pahu hoʻomau, ʻaʻole hiki ke loaʻa ka papa kuhikuhi waihona i nā hihia mua ʻelua, akā i nā hihia ʻelua hope e mālama ʻia ke komo. Eia nō naʻe, i ka hihia mua, hiki iā ʻoe ke komo i ka waihona ma o kahi host Docker e holo ana ma kahi mīkini virtual. I ka lua o ka hihia, ʻaʻole e nalowale ka ʻikepili, no ka mea ua hana ʻoe i kahi waihona Explicit.
Inā hāʻule ka mea hoʻokipa, ʻaʻole i loaʻa ka papa kuhikuhi waihona i nā hihia mua ʻekolu; i ka hihia hope, ʻaʻole i hoʻopau ʻia ka pilina me ka waihona. ʻO ka mea hope loa, ua kāpae ʻia ka hana kaʻana like no ka mālama ʻana i ka hihia mua a hiki i ke koena. I ka lua o ka hihia, hiki iā ʻoe ke kaʻana like i ka waihona ma muli o ke kākoʻo ʻana o kāu waihona i ka waiho ʻana a i ʻole. I ka hihia o Per-Host, hiki ke hāʻawi ʻia ka ʻikepili ma kahi pūʻali i hāʻawi ʻia, a no ka multihost ua hāʻawi ʻia e ka hoʻonui ʻana i ka cluster.
Pono e noʻonoʻo ʻia kēia i ka hana ʻana i nā ipu stateful. ʻO kahi mea hana Docker ʻē aʻe ʻo ia ka Volume plugin, e hana ana ma ke kumumanaʻo o "nā pahu i kēia manawa, akā pono e hoʻololi ʻia." Ke hoʻomaka ʻoe i kahi pahu Docker, ʻōlelo ʻia, "E, ke hoʻomaka ʻoe i kahi pahu me kahi waihona, hiki iā ʻoe ke mālama i kāu ʻikepili i kēia pahu!" ʻO kēia ka hiʻohiʻona paʻamau, akā hiki iā ʻoe ke hoʻololi. ʻAe kēia plugin iā ʻoe e hoʻohana i kahi kaʻa pūnaewele a i ʻole kekahi mea like ma kahi o kahi waihona waihona. Loaʻa iā ia kahi mea hoʻokele paʻamau no ka mālama ʻana i ka host-based a ʻae i ka hoʻohui ʻana i ka pahu me nā ʻōnaehana mālama waho e like me Amazon EBS, Azure Storage a me GCE Persistent disks.
Hōʻike ka slide aʻe i ka hoʻolālā o ka Docker Volume plugin.
ʻO ke kala polū e hōʻike ana i ka mea kūʻai aku Docker e pili ana me ka host Docker blue, nona kahi mīkini mālama kūloko e hāʻawi iā ʻoe i nā ipu no ka mālama ʻana i ka ʻikepili. Hōʻike ʻo Green i ka Plugin Client a me Plugin Daemon, i pili pū ʻia me ka host. Hāʻawi lākou i ka manawa e mālama ai i ka ʻikepili i ka mālama pūnaewele o ke ʻano o Storage Backend āu e pono ai.
Hiki ke hoʻohana ʻia ka plugin Docker Volume me ka waiho ʻana o Portworx. ʻO ka module PX-Dev maoli kahi pahu āu e holo ai e pili ana i kāu host Docker a hiki iā ʻoe ke mālama maʻalahi i ka ʻikepili ma Amazon EBS.
Hāʻawi ka mea kūʻai aku ʻo Portworx iā ʻoe e nānā i ke kūlana o nā pahu mālama like ʻole i pili i kāu mea hoʻokipa. Inā ʻoe e kipa i kaʻu blog, hiki iā ʻoe ke heluhelu pehea e hana ai i ka hapa nui o Portworx me Docker.
ʻO ka manaʻo o ka mālama ʻana ma Kubernetes e like me Docker a hōʻike ʻia e nā papa kuhikuhi i hiki ke loaʻa i kāu pahu i loko o kahi pod. Kūʻokoʻa lākou i ke ola o kekahi pahu. ʻO nā ʻano waihona maʻamau i loaʻa ʻo hostPath, nfs, awsElasticBlockStore, a me gsePersistentDisk. E nānā kākou i ka hana ʻana o kēia mau hale kūʻai ma Kubernetes. ʻO ka maʻamau, ʻo ke kaʻina o ka hoʻopili ʻana iā lākou he 3 mau ʻanuʻu.
ʻO ka mea mua, ʻo kekahi ma ka ʻaoʻao pūnaewele, ʻo ka mea maʻamau he luna hoʻomalu, hāʻawi iā ʻoe i kahi mālama hoʻomau. Aia kahi faila hoʻonohonoho PersistentVolume pili no kēia. A laila, kākau ka mea hoʻomohala noi i kahi faila hoʻonohonoho i kapa ʻia ʻo PersistentVolumeClaim, a i ʻole kahi noi mālama mālama PVC, e ʻōlelo ana: "Ua loaʻa iaʻu he 50GB o ka waihona i hāʻawi ʻia, akā i mea e hoʻohana ai nā poʻe ʻē aʻe i kona hiki, ke haʻi aku nei au i kēia PVC i kēia manawa. pono 10 GB wale nō". ʻO ka mea hope loa, ʻo ke kolu o ka ʻanuʻu ʻo ia kāu noi e kau ʻia ma ke ʻano he waihona, a hoʻomaka ka hoʻohana ʻana i ka pod, a i ʻole replica set, a i ʻole kekahi mea like. He mea nui e hoʻomanaʻo i kēia kaʻina hana i nā ʻanuʻu 3 i ʻōlelo ʻia a hiki ke scalable.
Hōʻike ka paheʻe aʻe i ka Kubernetes Persistence Container o ka hoʻolālā AWS.
I loko o ka ʻāpana ʻāpana ʻeleʻele e hōʻike ana i ka pūʻulu Kubernetes, hoʻokahi kumu kumu a me ʻelua node limahana, i hōʻike ʻia i ka melemele. Loaʻa i kekahi o nā node limahana kahi pod ʻalani, waihona, kahi mea hoʻoponopono kope, a me kahi pahu Docker Couchbase ʻōmaʻomaʻo. I loko o ka pūʻulu, ma luna o nā nodes, kahi ʻāpana poni e hōʻike ana i ka Service hiki ke loaʻa mai waho. Paipai ʻia kēia hoʻolālā no ka mālama ʻana i ka ʻikepili ma ka hāmeʻa ponoʻī. Inā pono, hiki iaʻu ke mālama i kaʻu ʻikepili ma EBS ma waho o ka pūʻulu, e like me ka mea i hōʻike ʻia ma ka slide aʻe. He hiʻohiʻona maʻamau kēia no ka scaling, akā aia kahi ʻano kālā e noʻonoʻo ai i ka wā e hoʻohana ai - ʻo ka mālama ʻana i nā ʻikepili ma kahi o ka pūnaewele hiki ke ʻoi aku ke kumukūʻai ma mua o ka mea hoʻokipa. Ke koho ʻana i nā hoʻonā containerization, ʻo ia kekahi o nā manaʻo koʻikoʻi.
E like me Docker, hiki iā ʻoe ke hoʻohana i nā pahu Kubernetes mau me Portworx.
ʻO kēia ka mea i kapa ʻia ʻo Kubernetes 1.6 huaʻōlelo i kēia manawa he "StatefulSet" - kahi ala e hana ai me nā palapala Stateful e hana ana i nā hanana e pili ana i ka hoʻōki ʻana i ka Pod a me ka hana ʻana i ka Graceful Shutdown. I kā mākou hihia, ʻo ia mau noi he waihona. Ma kaʻu blog hiki iā ʻoe ke heluhelu pehea e hana ai i kahi StatefulSet ma Kubernetes me ka hoʻohana ʻana iā Portworx.
E kamaʻilio kākou e pili ana i ka ʻaoʻao hoʻomohala. E like me kaʻu i ʻōlelo ai, loaʻa iā Docker nā mana 2 - CE a me EE, i ka hihia mua ke kamaʻilio nei mākou e pili ana i kahi mana paʻa o ka Community Edition, i hoʻonui ʻia i hoʻokahi manawa i kēlā me kēia 3 mahina, ʻokoʻa i ka mana hou o ka mahina o EE. Hiki iā ʻoe ke hoʻoiho iā Docker no Mac, Linux a i ʻole Windows. Ke hoʻokomo ʻia, e hoʻohou hou ʻo Docker a maʻalahi loa e hoʻomaka.
No nā Kubernetes, makemake au i ka mana Minikube - he ala maikaʻi ia e hoʻomaka ai me ka paepae ma ka hana ʻana i kahi puʻupuʻu ma ka node hoʻokahi. No ka hana ʻana i nā pūʻulu o kekahi mau nodes, ʻoi aku ka laulā o ke koho o nā mana: ʻo ia nā kops, kube-aws (CoreOS+AWS), kube-up (kahiko). Inā ʻoe e ʻimi nei e hoʻohana i nā Kubernetes i hoʻokumu ʻia ma AWS, paipai wau e hui pū me ka AWS SIG, e hui ana ma ka pūnaewele i kēlā me kēia Pōʻalima a hoʻopuka i nā ʻano mea hoihoi e pili ana i ka hana ʻana me AWS Kubernetes.
E nānā kākou i ka hana ʻana o Rolling Update ma kēia mau kahua. Inā loaʻa kahi pūʻulu o nā nodes, a laila hoʻohana ia i kahi mana kikoʻī o ke kiʻi, no ka laʻana, WildFly:1. ʻO ka hoʻololi hou ʻana, ʻo ia hoʻi, ua hoʻololi ʻia ka mana kiʻi me kahi mea hou ma kēlā me kēia node, kekahi ma hope o kekahi.
No ka hana ʻana i kēia, hoʻohana wau i ke kauoha docker service update (inoa lawelawe), kahi aʻu e kuhikuhi ai i ka mana hou o ka WildFly: 2 kiʻi a me ke ʻano o ka hoʻoponopono hou-parallelism 2. ʻO ka helu 2 ʻo ia ka mea e hōʻano hou ka ʻōnaehana i nā kiʻi noi 2. i ka manawa like, a laila e hoʻopaneʻe ka 10 mau kekona i ka 10s, a laila e hoʻonui ʻia nā kiʻi 2 aʻe ma 2 mau nodes, etc. Hāʻawi ʻia kēia ʻōnaehana hoʻololi maʻalahi iā ʻoe ma ke ʻano o Docker.
Ma Kubernetes, hana ʻia kahi hana hou e like me kēia. Hoʻokumu ka replication controller rc i kahi o nā replicas o ka mana like, a ua hāʻawi ʻia kēlā me kēia pod i kēia webapp-rc me kahi lepili aia ma etcd. Ke makemake au i kahi pod, hoʻohana wau i ka Service Application e komo i ka waihona etcd, e hāʻawi iaʻu i ka pod me ka hoʻohana ʻana i ka lepili i kuhikuhi ʻia.
I kēia hihia, loaʻa iā mākou he 3 pods i ka Replication controller e holo ana i ka polokalamu WildFly version 1. I ka wā e hoʻonui ai i ke kua, ua hana ʻia kahi mea hoʻoponopono hou me ka inoa like a me ka index ma ka hopena - - xxxxx, kahi x he mau helu random, a me nā lepili like. I kēia manawa, ʻekolu pods ka Service Application me ka mana kahiko o ka noi a ʻekolu pods me ka mana hou i ka mana Replication hou. Ma hope o kēia, hoʻopau ʻia nā pods kahiko, kapa ʻia ka mea hoʻoponopono replication me nā pods hou a hoʻokomo ʻia i ka hana.
E neʻe kākou i ka nānā ʻana. He nui nā kauoha nānā i kūkulu ʻia ʻo Docker. No ka laʻana, ʻo ka docker container stats command line interface hiki iā ʻoe ke hōʻike i ka ʻike e pili ana i ke kūlana o nā ipu i ka console i kēlā me kēia kekona - ka hoʻohana ʻana i ka ʻōnaehana, ka hoʻohana ʻana i ka disk, ka ukana pūnaewele. Hāʻawi ka Docker Remote API i ka ʻikepili e pili ana i ke kamaʻilio ʻana o ka mea kūʻai aku me ke kikowaena. Hoʻohana ia i nā kauoha maʻalahi, akā hoʻokumu ʻia ma ka Docker REST API. I kēia hihia, ʻo nā huaʻōlelo REST, Flash, Remote ke ʻano like. Ke kamaʻilio ʻoe me ka mea hoʻokipa, he REST API ia. Hāʻawi ka Docker Remote API iā ʻoe e kiʻi i ka ʻike hou aku e pili ana i ka holo ʻana i nā ipu. Hōʻike kaʻu blog i nā kikoʻī o ka hoʻohana ʻana i kēia nānā ʻana me Windows Server.
ʻO ka nānā ʻana i nā hanana ʻōnaehana docker i ka wā e holo ana i kahi puʻupuʻu multi-host e hiki ai ke loaʻa ka ʻikepili e pili ana i kahi ulia host a i ʻole kahi pahu pahu ma kahi host kikoʻī, nā lawelawe scaling, a me nā mea like. E hoʻomaka ana me Docker 1.20, loaʻa iā Prometheus, kahi e hoʻopili ai i nā hopena i nā noi i loaʻa. Hiki iā ʻoe ke loaʻa nā metric ma o HTTP a hōʻike iā lākou ma nā dashboards.
ʻO kahi hiʻohiʻona ʻē aʻe he cAdvisor (pōkole no ke kākāʻōlelo pahu). Hoʻopili a hāʻawi ia i ka hoʻohana waiwai a me ka ʻikepili hana mai nā pahu holo, e hāʻawi ana i nā metric Prometheus ma waho o ka pahu. ʻO ka mea kūikawā e pili ana i kēia mea hana, ʻo ia wale nō ka hāʻawi ʻana i ka ʻikepili no nā kekona 60 hope loa. No laila, pono ʻoe e hōʻiliʻili i kēia ʻikepili a waiho i loko o kahi waihona i hiki iā ʻoe ke nānā i kahi kaʻina hana lōʻihi. Hiki ke hoʻohana ʻia e hōʻike i nā metric dashboard me ka hoʻohana ʻana iā Grafana a i ʻole Kibana. Loaʻa i kaʻu blog kahi wehewehe kikoʻī o ka hoʻohana ʻana i ka cAdvisor e nānā i nā pahu me ka hoʻohana ʻana i ka dashboard Kibana.
Hōʻike ka paheʻe aʻe i ke ʻano o ka hopena hopena Prometheus a me nā metric i loaʻa e hōʻike.
Ma ka ʻaoʻao hema ʻike ʻoe i nā metric no nā noi HTTP, nā pane, a me nā mea ʻē aʻe, ma ka ʻākau ka hōʻike kiʻi.
Hoʻokomo pū ʻia nā Kubernetes i nā mea hana nānā i kūkulu ʻia. Hōʻike kēia paheʻe i kahi puʻupuʻu maʻamau i loaʻa hoʻokahi haku a me ʻekolu node limahana.
Loaʻa i kēlā me kēia o nā node hana kahi cAdvisor i hoʻokuʻu ʻia. Eia hou, aia ʻo Heapster, kahi ʻōnaehana nānā hana a me ka ʻohi ʻana i nā metric i kūpono me Kubernetes version 1.0.6 a ʻoi aku. Hāʻawi ʻo Heapster iā ʻoe e hōʻiliʻili ʻaʻole wale i nā ana hana o nā haʻahaʻa hana, nā pods a me nā ipu, akā pū kekahi me nā hanana a me nā hōʻailona ʻē aʻe i hana ʻia e ka hui holoʻokoʻa. No ka hōʻiliʻili ʻana i ka ʻikepili, kamaʻilio ʻo ia i kēlā me kēia pod o Kubelet, mālama ʻokoʻa i ka ʻike ma ka waihona InfluxDB, a hoʻopuka iā ia ma ke ʻano he ana i ka dashboard Grafana. Eia naʻe, e hoʻomanaʻo inā ʻoe e hoʻohana ana i ka miniKube, ʻaʻole i loaʻa kēia hiʻohiʻona ma ka paʻamau, no laila pono ʻoe e hoʻohana i nā addons no ka nānā ʻana. No laila, pili ia i kahi e holo ai ʻoe i nā ipu a me nā mea hana nānā e hiki ai iā ʻoe ke hoʻohana ma ka paʻamau a pono ʻoe e hoʻokomo i nā mea hoʻohui kaʻawale.
Hōʻike ka paheʻe hope i nā dashboards Grafana e hōʻike ana i ke kūlana holo o kaʻu mau ipu. Nui nā ʻikepili hoihoi ma aneʻi. ʻO kaʻoiaʻiʻo, he nui nā mea hoʻokele kaʻina hana Docker a me Kubernetes, e like me SysDig, DataDog, NewRelic. ʻO kekahi o lākou he 30 mau makahiki hoʻāʻo manuahi, no laila hiki iā ʻoe ke hoʻāʻo a loaʻa ka mea kūpono iā ʻoe. ʻO wau iho, makemake wau e hoʻohana iā SysDig a me NewRelic, e hoʻohui maikaʻi me Kubernetes. Aia nā mea hana e hoʻohui like i loko o nā kahua Docker a me Kubernetes.
Kekahi mau hoʻolaha 🙂
Mahalo no kou noho pū ʻana me mākou. Makemake ʻoe i kā mākou ʻatikala? Makemake ʻoe e ʻike i nā mea hoihoi hou aʻe? E kākoʻo iā mākou ma ke kau ʻana i kahi kauoha a i ʻole ka ʻōlelo ʻana i nā hoaaloha, , he analogue kū hoʻokahi o nā kikowaena hoʻokomo-level, i hana ʻia e mākou no ʻoe: (loaʻa me RAID1 a me RAID10, a hiki i 24 cores a hiki i 40GB DDR4).
ʻO Dell R730xd 2 mau manawa maʻalahi ma Equinix Tier IV kikowaena data ma Amsterdam? Eia wale nō ma Netherlands! Dell R420 - 2x E5-2430 2.2Ghz 6C 128GB DDR3 2x960GB SSD 1Gbps 100TB - mai $99! Heluhelu e pili ana
Source: www.habr.com