opensource/cloudpods
3
0
Fork 0
mirror of https://github.com/yunionio/cloudpods.git synced 2026-10-31 02:38:07 +08:00
Code Issues Packages Projects Releases Wiki Activity
20,871 Commits 30 Branches 712 Tags
9d85cfcd5d1c98783cd98d6a00a53dff376cb695
Commit Graph
20871 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Zexi Li 9d85cfcd5d fix(llm): 更新 model catalog 默认 URL 与 webtop 镜像同步 (#24974) 
将 ModelCatalogURL 默认路径改为 model-catalog.yaml；
sync_llm_images 改为同步 fedora-xfce webtop 并注释其它候选镜像。
 2026-06-04 18:51:20 +08:00
Zexi Li dd719de15c fix(llm): container desktop (#24969) 2026-06-04 11:51:29 +08:00
Zexi Li 40beca15e7 fix(monitor): alert_state of resource metrics api (#24965) 2026-06-03 17:15:44 +08:00
屈轩 fac826012e fix(region): extend external_id to 512 (#24962) 2026-06-03 15:27:58 +08:00
Zexi Li df815ad02b feat(llm): container desktop (#24960) 2026-06-03 13:24:39 +08:00
屈轩 71f8c3004c fix(region): add cpu model nic bandwidth for server sku (#24953) 2026-06-03 10:59:48 +08:00
Jian QiuandQiu Jian de012133d6 fix: compatibility issue with vms of empty hypervisor (#24958) 
Co-authored-by: Qiu Jian <qiujian@yunionyun.com>
 2026-06-03 09:57:53 +08:00
wanyaoqi d6478049a1 fix(mps): parse pci addr convert to lower case addr (#24943) 2026-06-02 20:03:02 +08:00
wanyaoqi 17e67af77c fix(host): check cluster is valid before do qemuimg convert (#24947) 2026-06-02 20:02:04 +08:00
屈轩 69d3a2d894 fix(region): filter zone region by read only (#24950) 2026-06-02 11:15:20 +08:00
屈轩 b8835bb333 fix(region): support skip server sync by tags (#24948) 2026-06-02 11:05:10 +08:00
屈轩 d20f102508 fix(region): avoid create huawei windows server fail (#24939) 2026-06-01 15:52:44 +08:00
屈轩 47fab40fc0 fix(region): init cloudpods host type (#24933) 2026-05-31 14:17:15 +08:00
屈轩 2346d2691a fix(region): init external project (#24935) 2026-05-31 12:45:57 +08:00
cwz_eikoh 8a676e6b4c Feat/auto gpu ultilization (#24924) 
* fix(llm): support hf import llm_sku

* feat(llm): auto-gpu-memory-utilization
 2026-05-29 17:45:35 +08:00
屈轩 83a32e14f4 fix(host-deployer): lvm lock (#24926) 2026-05-29 15:01:15 +08:00
wanyaoqi 31e1ea64d9 fix(host): guest sync override cpu_numa_pin before update (#24914) 2026-05-28 19:58:45 +08:00
屈轩 cdc399e6d7 fix(region): add notices sync interval (#24922) 2026-05-28 19:47:51 +08:00
屈轩 6a1c3ea3ad fix(region): remove account id (#24919) 2026-05-28 17:40:28 +08:00
Zexi Li 08bcd59231 fix(ansibleserver): allow CreatorMark to be nullable in AnsiblePlaybookV2 (#24918) 2026-05-28 16:38:19 +08:00
cwz_eikoh 76e30df4f5 Feature/download progress (#24885) 
* feat(llm): support download progress when model import

* fix(llm): unsupported catalog source \"ollama\" when model_set import llm_sku
 2026-05-27 13:20:47 +08:00
Zexi Li 982807b66b feat(monitor): add ResourceMetricsDefaultInterval option (#24908) 2026-05-27 11:50:48 +08:00
屈轩 ce4ef5af78 fix: upgrade go 1.26 (#24904) 2026-05-27 11:22:07 +08:00
LiuChenandMrliuch 2ad20fff33 fix: throttle image cache progress callback to avoid download bottleneck (#24898) 
When using S3/MinIO backend, the progress callback in AcquireImage was
called synchronously on every 4KB read chunk (via StreamPipe2). Each
call invokes UpdateServerProgress which makes a blocking HTTP PUT to the
compute API (~50ms), capping download speed at ~0.65 Mbps regardless of
available bandwidth.

Fix: throttle the callback to fire at most once every 5 seconds and make
the HTTP call asynchronous (goroutine), matching the behavior of v3.11.x
which used a 1-second ticker with async callbacks.

Fixes #24897

Co-authored-by: Mrliuch <liu15094534492@gmail.com>
 2026-05-26 15:09:46 +08:00
屈轩 52f84f0cd3 fix(region): skip useless addition network wire create (#24893) 2026-05-25 19:45:43 +08:00
Zexi Li 66ae3c91a7 fix(host): aggregate same-type container devices for env config (#24892) 
When multiple NVIDIA GPUs were attached to a container, each device was
passed individually to GetContainerExtraConfigures, producing one
NVIDIA_VISIBLE_DEVICES env per GPU. Duplicate env keys are collapsed by
the runtime to the last value, so the container only saw one GPU even
though the spec listed all of them.

Group RestDevs by DeviceType and invoke GetContainerExtraConfigures
once per type so all GPU UUIDs are joined into a single
NVIDIA_VISIBLE_DEVICES value.
 2026-05-25 17:30:10 +08:00
Zexi Li 9ce9949b21 feat(llm,scheduler): vram calculation (#24889) 2026-05-25 16:27:28 +08:00
Jian QiuandQiu Jian 53da0f15fd fix: disable s3 direct download default (#24886) 
Co-authored-by: Qiu Jian <qiujian@yunionyun.com>
 2026-05-25 10:03:54 +08:00
cwz_eikoh 588c38ec74 feat(llm): auto append backend_parameters in vllm & sglang (#24876) 2026-05-22 17:01:54 +08:00
Zexi Li ae431939fd fix(region): prefer check_code series for lb backend check_status (#24878) 2026-05-22 14:17:39 +08:00
Jian QiuandQiu Jian 2ecbbf31d9 fix: set bridge up after created (#24874) 
Co-authored-by: Qiu Jian <qiujian@yunionyun.com>
 2026-05-21 20:31:03 +08:00
wanyaoqi fcf79ceda8 fix(scheduler): set options ContainerNumaAllocate default true (#24869) 2026-05-21 19:31:44 +08:00
Zexi Li 07cb7a5a6f feat(region,host): upload nvidia gpu memory size (#24868) 2026-05-21 16:20:40 +08:00
cwz_eikohandZexi Li b0cf03c486 feat(hermes): connect vllm/sglang/ollama (#24842) 
Co-authored-by: Zexi Li <zexi.li@icloud.com>
 2026-05-21 11:41:01 +08:00
cwz_eikoh ff6d21e6ca feat(llm): create llm_sku from llm_model_set (#24855) 2026-05-21 11:40:23 +08:00
cwz_eikoh f8ee734d8f feat(llm): Implement resumable downloads for models from Hugging Face (#24843) 2026-05-21 10:48:27 +08:00
Zexi Li 389583d36c fix(region): ensure lb backend check_status returned when check_code missing (#24856) 2026-05-21 10:34:02 +08:00
wanyaoqi 0be8e92d17 fix(host): remove cgroup root path check quotation marks (#24850) 2026-05-21 10:30:14 +08:00
Zexi Li be90e452dc feat(llm): llm deployemnt (#24848) 2026-05-20 11:06:23 +08:00
Jian Qiu 6649b066ba Revert "fix: revert host masquerade setting (#24827)" (#24846) 
This reverts commit 81149c89f6.
 2026-05-19 21:01:42 +08:00
Zexi Li 02a83dc4ab fix(region): query metrics from monitor service (#24840) 2026-05-19 18:02:41 +08:00
Zexi Li 83d434bceb fix(llm): persistent /config dir of hermes agent (#24835) 2026-05-18 14:57:39 +08:00
Zexi Li 14b3048e61 feat(climc): add generic feature-config command (#24830) 
Add a generic `feature-config <feature> <on|off>` command alongside the
existing per-feature `feature-config-<name>` commands. A `--force` flag
skips the known-features validation so new features can be toggled
without a code change. Also adds "ai" to the known feature list.
 2026-05-18 14:28:46 +08:00
Jian QiuandQiu Jian 81149c89f6 fix: revert host masquerade setting (#24827) 
Co-authored-by: Qiu Jian <qiujian@yunionyun.com>
 2026-05-15 01:56:52 +08:00
cwz_eikoh 211619b685 feat(llm): support sglang (#24820) 2026-05-14 16:47:01 +08:00
Jian QiuandQiu Jian ab3c557878 fix: lbagent keepalived service run may exit (#24804) 
Co-authored-by: Qiu Jian <qiujian@yunionyun.com>
 2026-05-13 12:07:43 +08:00
屈轩 ead788f066 fix(region): support stop for change vm config (#24815) 2026-05-12 19:48:02 +08:00
屈轩 8c200905b2 fix(region): update gp3 disk size (#24808) 2026-05-12 17:27:52 +08:00
cwz_eikoh aec0118414 Feature(vllm): use entrypint instead of StartLLM (#24801) 
* feat(llm): use entrypoint instead of StartLLM

* feat(llm): adapt front-end in comfyui instant-model
 2026-05-12 10:43:05 +08:00
Zexi Li 6c4d68f3b8 fix(llm): honor SLLM device override when building container spec (#24799) 2026-05-11 20:06:35 +08:00