在 debian buster 服务器上，我有一个 apache2 Web 服务器，前面有 varnish 作为缓存，实际上 varnish 监听端口 80 并从监听端口 8080 的 apache 获取数据。
\n也许自从从stretch升级后，文件下载不再完成。
\n使用curl下载：
\n$ wget -v -t 1 https://myserver.org/myfile.zip\n--2020-07-10 19:52:34-- https://myserver.org/myfile.zip\nResolving myserver.org (myserver.org)... w.x.y.z\nConnecting to myserver.org (myserver.org)|w.x.y.z|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 333774113 (318M) [application/zip]\nSaving to: 'myfile.zip.13'\n\nmyfile.zip.13 1%[=> ] 4.43M 8.29MB/s in 0.5s \n\n2020-07-10 19:52:35 (8.29 MB/s) - Connection closed at byte 4649063. Giving up.\n每个请求的传输量都不同。
\n这是此请求的清漆日志：
\n* << Request >> 2104598 \n- Begin req 2104597 rxreq\n- Timestamp Start: 1594404000.669087 0.000000 0.000000\n- Timestamp Req: 1594404000.669087 0.000000 0.000000\n- ReqStart 127.0.0.1 40966 a0\n- ReqMethod GET\n- ReqURL /myfile.zip\n- ReqProtocol HTTP/1.1\n- ReqHeader User-Agent: Wget/1.19.4 (linux-gnu)\n- ReqHeader Accept: */*\n- ReqHeader Accept-Encoding: identity\n- ReqHeader Host: myserver.org\n- ReqHeader X-SSL: 1\n- ReqHeader X-Forwarded-Proto: https\n- ReqHeader X-Forwarded-For: a.b.c.d\n- ReqHeader Connection: close\n- ReqUnset X-Forwarded-For: a.b.c.d\n- ReqHeader X-Forwarded-For: a.b.c.d, 127.0.0.1\n- VCL_call RECV\n- ReqUnset Accept-Encoding: identity\n- VCL_return hash\n- VCL_call HASH\n- VCL_return lookup\n- VCL_call MISS\n- VCL_return fetch\n- Link bereq 2104599 fetch\n- Timestamp Fetch: 1594404000.669596 0.000509 0.000509\n- RespProtocol HTTP/1.1\n- RespStatus 200\n- RespReason OK\n- RespHeader Date: Fri, 10 Jul 2020 18:00:00 GMT\n- RespHeader Last-Modified: Sun, 28 Jun 2020 14:46:46 GMT\n- RespHeader ETag: "13e4fd21-5a92602e5bcc1"\n- RespHeader Content-Length: 333774113\n- RespHeader Content-Type: application/zip\n- RespHeader X-Varnish: 2104598\n- RespHeader Age: 0\n- RespHeader Via: 1.1 varnish (Varnish/6.1)\n- VCL_call DELIVER\n- RespHeader X-Cache: uncached\n- RespUnset Via: 1.1 varnish (Varnish/6.1)\n- RespUnset X-Varnish: 2104598\n- VCL_return deliver\n- Timestamp Process: 1594404000.669609 0.000521 0.000012\n- RespHeader Accept-Ranges: bytes\n- RespHeader Connection: close\n- Timestamp Resp: 1594404000.884468 0.215381 0.214859\n- ReqAcct 237 0 237 260 4649063 4649323\n- End\n这是后端交互（其他wget运行，在字节6291456处关闭）：
\n* << BeReq >> 2612416 \n- Begin bereq 2612415 fetch\n- Timestamp Start: 1594674394.093379 0.000000 0.000000\n- BereqMethod GET\n- BereqURL /myfile.zip\n- BereqProtocol HTTP/1.1\n- BereqHeader User-Agent: Wget/1.19.4 (linux-gnu)\n- BereqHeader Accept: */*\n- BereqHeader Host: myserver.org\n- BereqHeader X-SSL: 1\n- BereqHeader X-Forwarded-Proto: https\n- BereqHeader X-Forwarded-For: a.b.c.d, 127.0.0.1\n- BereqHeader Accept-Encoding: gzip\n- BereqHeader X-Varnish: 2612416\n- VCL_call BACKEND_FETCH\n- VCL_return fetch\n- BackendOpen 28 default 127.0.0.1 8080 127.0.0.1 50218\n- BackendStart 127.0.0.1 8080\n- Timestamp Bereq: 1594674394.093413 0.000034 0.000034\n- Timestamp Beresp: 1594674394.093710 0.000330 0.000297\n- BerespProtocol HTTP/1.1\n- BerespStatus 200\n- BerespReason OK\n- BerespHeader Date: Mon, 13 Jul 2020 21:06:34 GMT\n- BerespHeader Server: Apache\n- BerespHeader Last-Modified: Sun, 28 Jun 2020 14:46:46 GMT\n- BerespHeader ETag: "13e4fd21-5a92602e5bcc1"\n- BerespHeader Accept-Ranges: bytes\n- BerespHeader Content-Length: 333774113\n- BerespHeader Content-Type: application/zip\n- TTL RFC 120 10 0 1594674394 1594674394 1594674394 0 0 cacheable\n- VCL_call BACKEND_RESPONSE\n- BerespUnset Server: Apache\n- TTL VCL 1800 10 0 1594674394 cacheable\n- TTL VCL 1800 30 0 1594674394 cacheable\n- VCL_return deliver\n- Filters \n- Storage malloc s0\n- Fetch_Body 3 length stream\n- ExpKill LRU_Cand p=0x7f97b4a358c0 f=0x0 r=1\n- ExpKill LRU x=2631322\n- ExpKill LRU_Cand p=0x7f97efa67980 f=0x0 r=1\n- ExpKill LRU x=2284258\n- ExpKill LRU_Cand p=0x7f97b4a6b1c0 f=0x0 r=1\n- ExpKill LRU x=2631334\n- ExpKill LRU_Cand p=0x7f97efa67a40 f=0x0 r=1\n- ExpKill LRU x=2284266\n- ExpKill LRU_Cand p=0x7f97efa67b00 f=0x0 r=1\n- ExpKill LRU x=2284269\n- ExpKill LRU_Cand p=0x7f97d484c780 f=0x0 r=1\n- ExpKill LRU x=1667855\n- ExpKill LRU_Cand p=0x7f97d9af1bc0 f=0x0 r=1\n- ExpKill LRU x=2950859\n- ExpKill LRU_Cand p=0x7f97e7886040 f=0x0 r=1\n- ExpKill LRU x=2472240\n- ExpKill LRU_Cand p=0x7f97d9af1c80 f=0x0 r=1\n- ExpKill LRU x=2950862\n- ExpKill LRU_Cand p=0x7f97d484e640 f=0x0 r=1\n- ExpKill LRU x=1667862\n- ExpKill LRU_Cand p=0x7f97d484eac0 f=0x0 r=1\n- ExpKill LRU x=1667865\n- ExpKill LRU_Cand p=0x7f97d9af1e00 f=0x0 r=1\n- ExpKill LRU x=2950865\n- ExpKill LRU_Cand p=0x7f97d48d2180 f=0x0 r=1\n- ExpKill LRU x=1667871\n- ExpKill LRU_Cand p=0x7f97d9af2040 f=0x0 r=1\n- ExpKill LRU x=2950871\n- ExpKill LRU_Cand p=0x7f97d9af2100 f=0x0 r=1\n- ExpKill LRU x=2950876\n- ExpKill LRU_Cand p=0x7f97b4b11000 f=0x0 r=1\n- ExpKill LRU x=2631343\n- ExpKill LRU_Cand p=0x7f97efa67bc0 f=0x0 r=1\n- ExpKill LRU x=2284275\n- ExpKill LRU_Cand p=0x7f97efa67c80 f=0x0 r=1\n- ExpKill LRU x=2284278\n- ExpKill LRU_Cand p=0x7f97d48d3b00 f=0x0 r=1\n- ExpKill LRU x=1667880\n- ExpKill LRU_Cand p=0x7f97e7886100 f=0x0 r=1\n- ExpKill LRU x=2472249\n- ExpKill LRU_Cand p=0x7f97d9a3d3c0 f=0x0 r=1\n- ExpKill LRU x=2950879\n- ExpKill LRU_Cand p=0x7f97d9a3d900 f=0x0 r=1\n- ExpKill LRU x=2950886\n- ExpKill LRU_Cand p=0x7f97d48d4040 f=0x0 r=1\n- ExpKill LRU x=1667895\n- ExpKill LRU_Cand p=0x7f97e79d1e80 f=0x0 r=1\n- ExpKill LRU x=2631367\n- ExpKill LRU_Cand p=0x7f97d6bf6780 f=0x0 r=1\n- ExpKill LRU x=1205206\n- ExpKill LRU_Cand p=0x7f97cbb3c2c0 f=0x0 r=1\n- ExpKill LRU x=2325555\n- ExpKill LRU_Cand p=0x7f97d6bf6840 f=0x0 r=1\n- ExpKill LRU x=1205217\n- ExpKill LRU_Cand p=0x7f97e7886340 f=0x0 r=1\n- ExpKill LRU x=2472273\n- ExpKill LRU_Cand p=0x7f97cbb3c380 f=0x0 r=1\n- ExpKill LRU x=2325569\n- ExpKill LRU_Cand p=0x7f97cbb3c440 f=0x0 r=1\n- ExpKill LRU x=2325572\n- ExpKill LRU_Cand p=0x7f97d6bf69c0 f=0x0 r=1\n- ExpKill LRU x=1205230\n- ExpKill LRU_Cand p=0x7f97e7886400 f=0x0 r=1\n- ExpKill LRU x=2472276\n- ExpKill LRU_Cand p=0x7f97d6bf6b40 f=0x0 r=1\n- ExpKill LRU x=1205239\n- ExpKill LRU_Cand p=0x7f97b4b11480 f=0x0 r=1\n- ExpKill LRU x=2631405\n- ExpKill LRU_Cand p=0x7f97b4b11540 f=0x0 r=1\n- ExpKill LRU x=2631410\n- ExpKill LRU_Cand p=0x7f97e7886640 f=0x0 r=1\n- ExpKill LRU x=2472294\n- ExpKill LRU_Cand p=0x7f97b4b116c0 f=0x0 r=1\n- ExpKill LRU x=2631416\n- ExpKill LRU_Cand p=0x7f97e7886700 f=0x0 r=1\n- ExpKill LRU x=2472299\n- ExpKill LRU_Cand p=0x7f97b4b11780 f=0x0 r=1\n- ExpKill LRU x=2631419\n- ExpKill LRU_Cand p=0x7f97e78867c0 f=0x0 r=1\n- ExpKill LRU x=2472302\n- ExpKill LRU_Cand p=0x7f97e7886880 f=0x0 r=1\n- ExpKill LRU x=2472305\n- ExpKill LRU_Cand p=0x7f97b4b11b40 f=0x0 r=1\n- ExpKill LRU x=2631428\n- ExpKill LRU_Cand p=0x7f97e7886940 f=0x0 r=1\n- ExpKill LRU x=2472308\n- ExpKill LRU_Cand p=0x7f97e7886ac0 f=0x0 r=1\n- ExpKill LRU x=2472314\n- ExpKill LRU_Cand p=0x7f97b4b12080 f=0x0 r=1\n- ExpKill LRU x=2631434\n- ExpKill LRU_Cand p=0x7f97d6bf6c00 f=0x0 r=1\n- ExpKill LRU x=1205267\n- ExpKill LRU_Cand p=0x7f97e7886b80 f=0x0 r=1\n- ExpKill LRU x=2472348\n- ExpKill LRU_Cand p=0x7f97d6bf6cc0 f=0x0 r=1\n- ExpKill LRU x=1205274\n- ExpKill LRU_Cand p=0x7f97b4b12200 f=0x0 r=1\n- ExpKill LRU x=2631441\n- ExpKill LRU_Cand p=0x7f97e7886c40 f=0x0 r=1\n- ExpKill LRU x=2472360\n- ExpKill LRU reached nuke_limit\n- FetchError Could not get storage\n- BackendClose 28 default\n- BereqAcct 244 0 244 230 10764288 10764518\n- End \n我无法理解发生了什么事...
\n这是我的清漆的default.vcl：
\n# Marker to tell the VCL compiler that this VCL has been adapted to the\n# new 4.0 format.\nvcl 4.0;\n\n# Default backend definition. Set this to point to your content server.\nbackend default {\n .host = "127.0.0.1";\n .port = "8080";\n}\n\n# access control list for "purge": open to only localhost and other local nodes\nacl purge {\n "127.0.0.1";\n "localhost";\n}\n\n# vcl_recv is called whenever a request is received \nsub vcl_recv {\n # Serve objects up to 2 minutes past their expiry if the backend\n # is slow to respond.\n #set req.grace = 120s;\n if (req.http.X-Forwarded-Proto !~ "https")\n {\n unset req.http.X-Forwarded-For;\n set req.http.X-Forwarded-For = client.ip;\n }\n set req.backend_hint= default;\n\n # This uses the ACL action called "purge". Basically if a request to\n # PURGE the cache comes from anywhere other than localhost, ignore it.\n if (req.method == "PURGE") \n {\n if (!client.ip ~ purge)\n {return(synth(405,"Not allowed."));}\n #return(hash);}\n return(purge);\n }\n\n # Pass any requests that Varnish does not understand straight to the backend.\n if (req.method != "GET" && req.method != "HEAD" &&\n req.method != "PUT" && req.method != "POST" &&\n req.method != "TRACE" && req.method != "OPTIONS" &&\n req.method != "DELETE") \n {return(pipe);} /* Non-RFC2616 or CONNECT which is weird. */\n\n # Pass anything other than GET and HEAD directly.\n if (req.method != "GET" && req.method != "HEAD")\n {return(pass);} /* We only deal with GET and HEAD by default */\n\n # Pass requests from logged-in users directly.\n #if (req.http.Authorization || req.http.Cookie)\n # {return(pass);} /* Not cacheable by default */\n\n # Pass any requests with the "If-None-Match" header directly.\n if (req.http.If-None-Match)\n {return(pass);}\n\n # Force lookup if the request is a no-cache request from the client.\n # DISATTIVO PERCH\xc3\x89 ban RICHIEDE UNA COMPARAZIONE\n #if (req.http.Cache-Control ~ "no-cache")\n # {ban(req.url);}\n\n # normalize Accept-Encoding to reduce vary\n if (req.http.Accept-Encoding) {\n if (req.http.User-Agent ~ "MSIE 6") {\n unset req.http.Accept-Encoding;\n } elsif (req.http.Accept-Encoding ~ "gzip") {\n set req.http.Accept-Encoding = "gzip";\n } elsif (req.http.Accept-Encoding ~ "deflate") {\n set req.http.Accept-Encoding = "deflate";\n } else {\n unset req.http.Accept-Encoding;\n }\n }\n\n\n # remove the cookies form static files in order to serve them from cache\n if (req.url ~ "(?i)\\.(jpeg|jpg|png|gif|ico|webp|js|css|txt|pdf|gz|zip|lzma|bz2|tgz|tbz|html|htm)$") {\n unset req.http.cookie;\n }\n\n # Do not cache HTTP authentication and HTTP Cookie\n if (req.http.Authorization || req.http.Cookie) {\n # Not cacheable by default\n return (pass);\n }\n\n return(hash);\n #return(pass);\n}\n\nsub vcl_pipe {\n # Note that only the first request to the backend will have\n # X-Forwarded-For set. If you use X-Forwarded-For and want to\n # have it set for all requests, make sure to have:\n # set req.http.connection = "close";\n \n # This is otherwise not necessary if you do not do any request rewriting.\n \n set req.http.connection = "close";\n}\n\n# Called if the cache has a copy of the page.\nsub vcl_hit {\n if (req.method == "PURGE") \n {ban(req.url);\n return(synth(200,"Purged"));}\n \n if (!obj.ttl > 0s)\n {return(pass);}\n}\n \n# Called if the cache does not have a copy of the page.\nsub vcl_miss {\n if (req.method == "PURGE") \n {return(synth(200,"Not in cache"));}\n}\n\n# Called after a document has been successfully retrieved from the backend.\nsub vcl_backend_response {\n \n # Remove some headers we never want to see\n unset beresp.http.Server;\n unset beresp.http.X-Powered-By;\n\n # For static content strip all backend cookies\n if (bereq.url ~ "\\.(css|js|png|gif|jp(e?)g)|swf|ico") {\n unset beresp.http.cookie;\n }\n # Don't store backend\n if (bereq.url ~ "wp-(login|admin)" || bereq.url ~ "preview=true") {\n set beresp.uncacheable = true;\n set beresp.ttl = 30s;\n return (deliver);\n }\n\n\n # don't cache response to posted requests or those with basic auth\n if ( bereq.method == "POST" || bereq.http.Authorization ) {\n set beresp.uncacheable = true;\n set beresp.ttl = 120s;\n return (deliver);\n }\n\n # don't cache search results\n if ( bereq.url ~ "\\?s=" ){\n set beresp.uncacheable = true;\n set beresp.ttl = 120s;\n return (deliver);\n }\n\n # only cache status ok\n if ( beresp.status != 200 ) {\n set beresp.uncacheable = true;\n set beresp.ttl = 120s;\n return (deliver);\n }\n\n # A TTL of 2h\n set beresp.ttl = 30m;\n # Define the default grace period to serve cached content\n set beresp.grace = 30s;\n return (deliver);\n}\n\nsub vcl_deliver {\n if (obj.hits > 0) {\n set resp.http.X-Cache = "cached";\n } else {\n set resp.http.x-Cache = "uncached";\n }\n\n # Remove some headers: PHP version\n unset resp.http.X-Powered-By;\n\n # Remove some headers: Apache version & OS\n unset resp.http.Server;\n\n # Remove some heanders: Varnish\n unset resp.http.Via;\n unset resp.http.X-Varnish;\n \n return(deliver);\n}\n\n#sub vcl_fetch {\n# if (req.url ~ "^/piwik") {\n# return(hit_for_pass);\n# }\n#}\n根据后端varnishlog信息，我可以得出结论，您的存储空间不足。
这些是确认这一点的日志项：
- ExpKill LRU reached nuke_limit
- FetchError Could not get storage
当缓存已满并且 Varnish 想要在缓存中存储另一个对象时，LRU（最近最少使用）机制就会启动。
这意味着它开始在缓存中查找最近最少使用的项目，并将它们从缓存中逐出。Varnish 会多次执行此操作，直到达到默认nuke_limit设置为50的值。
在这种情况下，Varnish将尝试删除最近最少使用的50个对象，以便将新对象放入缓存中。
我的假设是，50 个最近最少使用的对象的大小非常小，并且释放的空间与新对象的大小要求不匹配。
以下是您的一些选择：
nuke_limit使用运行时选项增加参数-p（-p nuke_limit=100例如）| 归档时间： |
|
| 查看次数： |
1321 次 |
| 最近记录： |