HttpConnectorServlet
The HttpConnectorServlet forwards HTTP requests to content providers using HTTP over TCP/IP. It supports load balancing and special handling of proxy-unaware content providers (content rewriting for HTML) as well as security features (cookie management). All its configuration attributes may also be used for the HttpsConnectorServlet.
ch::nevis::isiweb4::servlet::connector::http::HttpConnectorServlet
libHttpConnectorServlet.so.1
Profiles
JavaScript-Parsing
Regular expressions for some common JavaScript constructs containing URLs.
Configuration
InetAddress
Type: address list
Usage Constraints: required, basic connectivity
Sets a whitespace-separated list of INET addresses (<host>:<port>
, where <host>
may be an IP address or a host name). All backends must be on a single line and separated with blank spaces.
Failsafeness will work only if several server names are configured. The first server name is used as the ‘primary’ server (the one where all requests are sent to if load balancing is not enabled).
It is possible to define load balancing groups. Each group will be |
separated. For example: ip1:p1 ip2:p2 | ip3:p3 ip4:p4
.
In such a load balancing setup, the selection of the backend server depends on the following criteria:
- The backend server must be part of the defined load balancing group (in the above example, this is the group with servers ip1 and ip2);
- The backend server must not be marked as "bad".
The proxy makes its final selection of the backend server from among the available backend servers according to the Round Robin principle.
The proxy then stores the assignment of the client session to the selected backend server, thus binding this frontend/backend to the session. All further requests within this session are now sent to the same backend server.
In the case that the backend server has a problem, the proxy returns the request to the user together with an HTTP code 502
.
Also, the proxy removes the assigned backend server from the session.
In place of the old server, the proxy selects a new backend server from the same load balancing group for all further user requests within this session.
All the time, the session remains active and valid.
If there is no functioning backend server available in the first load balancing group, the proxy selects an appropriate backend server from the second load balancing group, according to the previously described selection criteria. In our example, the second group contains the servers ip3 and ip4.
You can also set a dynamic address with the following syntax:
<source>:<name>;[<source>:<name>;]...
with:
<source>
:ENV
,AUTH
,HEADER
,PARAM
, orSESSION
;<name>
: the name in the given source.
For example SESSION:host;
will get the host to connect to from the attribute host
in the session. You can set exactly one dynamic address (without any other address).
If IP addresses are used, Hostname
may need to be set.
You also can specify a YAML file, for example file:///var/opt/nevisproxy/<your instance name>/work/WEB-INF/connection.yml
, which you can change on the fly.
Read more about YAML configuration in the chapter below. The file-modification of the YAML file will be checked in the interval configured under periodicity
in the Timer
section in the file navajo.xml.
MappingType
Type: enum
Possible values: requesturi
, pathinfo
Usage Constraints: optional, basic connectivity
Default: requesturi
Configures the part of request path that will be sent to the content provider:
requesturi
: the complete URI of the incoming request will be sent to the content provider;pathinfo
: the pathinfo will be sent to the content provider.
requesturi = servlet mapping + pathinfo
AutoRewrite
Type: string array
Usage Constraints: optional, troubleshooting
Default: header
This attribute configures rewriting policies for non-proxy aware content in the HTTP/HTML response. It may be set to a newline or white space separated list of the following values:
redirect
: the HTTP headerLocation
will be re-writtencookie
: the HTTP headerSet-Cookie
will be rewritten (path information)link
: the HTTP headerLink
will be re-writtenbody
: HTML tags will be rewritten only in response to bodies of typetext/html
header
: alias for+redirect +cookie +link
all
: alias for+header +body
none
: neither the headers nor the body of the response are rewritten.
Each value can be prefixed either with a +
or a -
. No prefix is handled like a +
. If it is prefixed with a -
, then the correspondent value will not be overwritten.
For example, to rewrite all header except the link-header, set it to this value: header -link
.
Rewriting of the response body may only be omitted if the content provider is proxy aware.
See HostName
attribute for more information about content rewriting.
If cookie
is enabled for AutoRewrite
and the mappingType
is pathinfo
, the path of the cookie is rewritten in the same way that an absolute URI is written in the response body. In other words, the path will be prefixed with the servlet-path.
RequestUri.AdaptCookiePath
Type: boolean
Usage Constraints: optional
Default: false
This parameter is only evaluated if the MappingType
is requesturi
and AutoRewrite
contains cookie
rewriting.
If true
, the path of the cookies sent by the backend are set to the servlet-path if the backend did not send any path or if the path was set to '/'.
Avoid setting this parameter to true
if the same HttpConnectorServlet is mapped to several paths, unless you want the cookies only be visible for the path the HttpConnectorServlet is mapped to.
Avoid setting this parameter to true
if a UrlEncryptionFilter is configured on the same path. This is because on the client side the path will not match the encrypted URL.
URIPrefix
Type: string (absolute URI)
Usage Constraints: optional, advanced
This attribute configures the entry point of the content provider.
The behavior of this attribute depends on the attribute MappingType
:
pathInfo
: the configuredURIPrefix
will be prepended to the pathInforequestURI
: the configuredURIPrefix
will be prepended to the requestURI
requestURI = servlet mapping + pathinfo
For a more detailed description, see chapter Managing the URL namespace.
URIPrefix.trailingSlashRedirect
Type: boolean
Usage Constraints: optional, advanced
Default: false
Defines how to handle the URI /<URIPrefix>
. If set to true
, it will be treated like /<URIPrefix>/
.
CheckForMultipleIpAddresses
Type: boolean
Usage Constraints: optional
Default: false
If set to true
, all available IP addresses are requested on startup, and each IP address is handled as a separate host. If loadbalancing
is set to true
, then load balancing is done over those hosts as well.
Limitations:
- The DNS lookup is done when the proxy starts, and if the number of IP addresses changes sometime later, only the number we had at startup is considered. Assume we have 3 IP addresses at startup. If later we have 4, only the first 3 is used. If later we have 2 of them, one of them is used twice.
- Neither dynamic, nor YAML file-based
InetAddress
is supported.
CookieManager
Type: string array
Usage Constraints: optional, basic
Default: retain:^JSESSIONID$ allow:^.*$
This attribute configures the caching of cookies coming from the content provider. A cookie cached by the CookieManager is not forwarded to the browser, but merged into each incoming request before propagated to a content provider.
You set the mode of the CookieManager as follows: <rule>:<regexp>
The following rules are allowed:
store
: The CookieManager stores the cookie sent by the content provider within nevisProxy (the client does never receive the cookie and cannot send one either).store
is a suitable option if the connector is used for unprotected and protected content at the same time.retain
: If a session exists, the CookieManager stores the specified cookie (store
). Otherwise, it sends the cookie to the client (allow
).block
: The CookieManager blocks (that is, ignores) any cookie sent by the content provider or the client.allow
: The CookieManager allows the cookie sent by the content provider or client to pass through if there is no matchingstore
orblock
rule.
The regular expression is optional. If not configured, all cookies match. The default regex-type is PCRE(da)
. See also Regular expressions.
Examples:
store:^JSESSIONID$
store:^Navajo$
allow:^.*Language$
block:^.*Session.*$
Cookies that do not match any rule are blocked.
Use the following configuration to disable the CookieManager:
<param-value>off</param-value>
A disabled CookieManager prevents the logout call from sending any cookie to the content provider.
To find out which cookies are set by content providers, either:
- enable cookie warnings in your browser, or
- enable NavajoOp tracing (NavajoOp=1) and monitor
Set-Cookie
HTTP headers.
CookieManager.CookieBinding
Type: enum
Possible values: none
, resource
, mapping
Usage Constraints: optional, advanced
Default: none
Defines the scope of a stored cookie:
none
: cookie is used for each individual access to the servlet instance.resource
: Like none, but if multiple resources are configured for load balancing, the cookie is not shared between the resources.mapping
: All mappings of this servlet have their own set of cookies.
CookieManager.DefaultCookiePath
Type: string
Usage Constraints: optional, advanced
Default: /
The cookie path to set if the backend sets no path.
LogoutURI
Type: string array
Usage Constraints: optional, basic feature
Default: off
This attribute allows notifying the content provider about the termination of a session, to also terminate the application session.
The syntax of this attribute is:
<uri>(:<header>)*
For example:
/logout/path:User-Agent:Date
<uri>
: Describes the URI to be used for the notification call (through a GET request).<header>
: Specifies the names of the header fields to be delegated with the termination request (such as the header fieldsUser-Agent
andDate
in the previous example). The<header>
part is optional. If an optional header does not have a value, the header will not be sent with the request.
The session listener is required to trigger the notification when a session termination occurs. For more information about the session listener, see Session listener.
LogoutURI.Interception
Type: string
Usage Constraints: optional, advanced
Default: off
Configures the behaviour when a client requests the URI configured in the LogoutURI
parameter. The following actions can be taken upon such a request:
<url>
: An URL to redirect to.auto
: A redirect to the same location as requested will be triggered with an appendedlogout
url parameter.off
: Default, no specific action will be taken and the request will be forwarded to the backend server.
LogoutURI.NumberOfThreads
Type: integer
Usage Constraints: optional, advanced, min: 0
, max: 100
Default: 0
Number of threads to start to send the logout-call in separate threads. This will avoid to block the session-reaper.
LogoutURI.MaxNumberOfQueuedLogoutRequests
Type: integer
Usage Constraints: optional, advanced, min: 1
, max: 100000
Default: 1000
Maximal number of queued logouturi-requests. This value will only be used if LogoutURI.NumberOfThreads
is greater than 0
.
If this limit is reached, all following logouturi-requests will either be ignored (LogoutURI.Blocking=false
), or will be called in blocking mode (LogoutURI.Blocking=true
).
LogoutURI.Blocking
Type: boolean
Usage Constraints: optional, advanced
Default: true
If the limit set by LogoutURI.MaxNumberOfQueuedLogoutRequests
is reached, all following logouturi-requests will either be ignored (LogoutURI.Blocking=false
), or will be called in blocking mode (LogoutURI.Blocking=true
).
LoadBalancing
Type: boolean
Usage Constraints: optional, basic feature
Default: false
Activates load balancing.
Loadbalancing will work only if several server names are configured in InetAddress.
Loadbalancing is sticky, based on client session. To change this, set ResourceManager
accordingly.
DisableBindingStatusCode
Type: string
Usage Constraints: optional, advanced, conditions are supported
Default: not set, failover disabled
This attribute enables failover and defines the status code(s) for the content provider that failed. It is not set by default.
The connector first returns the response with the special status code before it removes the binding (from the HttpConnectorServlet configuration).
ResourceManager
Type: string
Usage Constraints: optional, advanced
Default: ch::nevis::isiweb4::servlet::connector::base::SessionResourceManager
Configures the classname of the used resource manager.
The following resource managers are available:
ch::nevis::isiweb4::servlet::connector::base::SessionResourceManager
: Per-session round-robin load balancing (session sticky).ch::nevis::isiweb4::servlet::connector::-base::StatelessResourceManager
: Per-request round-robin load balancing (no stickiness).
See also Fail-safety and load balancing.
ResourceHint
Type: string
Usage Constraints: optional, advanced
This attribute defines a numeric value, which is used as an index of a backend resource. Indexing starts at 1
.
Thus, this attribute allows direct access to a specific backend resource in case load balancing is enabled.
The selected backend resource is made persistent with the resource manager. You define the resource manager with the ResourceManager
attribute.
It is not possible to use this attribute together with a stateless resource manager (StatelessResourceManager
, see also the ResourceManager
attribute).
ResourceHint.Sticky
Type: boolean
Usage Constraints: optional, advanced
Default: false
Makes the backend selected with ResourceHint
in a load-balanced group sticky.
ResourceHint.RemoveFromQuery
Type: boolean
Usage Constraints: optional
Default: false
If this attribute is set to true
, the system removes the resource hint from the query that is sent to the backend.
ResourceManager.RetryTimeout
Type: integer [sec]
Usage Constraints: optional, advanced
Default: 10
; and if only one backend is set, and CheckForMultipleIPAddresses
is false
, the default is 0
Configures the retry timeout of the resource manager. Raise this timeout, if probing the backend takes a very long time (for example the server is stopped, network infrastructure does not immediately signal an "unreachable destination" and the ConnectTimeout cannot be lowered to, for example, less than 2 seconds). Disable it (by setting it to 0
) to avoid subsequent 502 errors, when the first was caused by a long-running transaction (for example generating a report) but the backend is not generally affected by performance problems.
ResourceManager.RetryBindingTimeout
Type: integer [sec]
Usage Constraints: optional, advanced
Default: 7200
Defines the time after which the connector will retry to bind the resource to a session.
ResourceManager.WatchURI
Type: string
Usage Constraints: optional, advanced
Defines a URI to call to check if a service is available or not. If not configured, only a TCP connect is done to check availability of a backend.
If a URIPrefix
is set, this prefix will be prepended to the given ResourceManager.WatchURI
.
ResourceManager.WatchStatusCode
Type: integer
Usage Constraints: optional, advanced, min: 100
, max: 600
The expected HTTP status code when the ResourceManager.WatchURI
is called. If the status code does not match the one configured, the backend is considered as unavailable.
ResourceManager.WatchTimeout
Type: integer [msec]
Usage Constraints: optional, advanced, min: 0
, max: 3600000
The request timeout for the call to ResourceManager.WatchURI
. If the backend did not answer within the configured timeout, the backend will be considered as unavailable.
ResourceManager.NoSessionCookie
See ResourceManager.NoSessionCookie.*
.
ResourceManager.NoSessionCookie.*
The attributes ResourceManager.NoSessionCookie
and ResourceManager.NoSessionCookie.*
define the NoSessionCookie feature.
- For a description of the NoSessionCookie feature, see NoSessionCookie.
- For a description of the
ResourceManager.NoSessionCookie.*
attributes, see NoSessionCookie configuration.
ResourceManager.FailoverThresholdCount
Type: integer
Usage Constraints: optional, advanced
Default: 1
The number of requests that shall need to fail before a particular backend is considered defective. Increase this to prevent a failover if an otherwise healthy backend occasionally fails to satisfy isolated requests.
The threshold setting is shared by all the resources configured for the servlet, the actual failure counters are maintained independently per backend.
Until the failure counter reaches the configured threshold, the backend continues to be considered available.
The configured ResourceManager will only proceed to consider other resources once the threshold has been reached.
If a ResourceManager.WatchURI
is configured, the ResourceManager.FailoverThresholdCount
will be ignored.
ResourceManager.FailoverThresholdCountLifeTime
Type: integer [sec]
Usage Constraints: optional, advanced, min: 0
, max: 604800
Default: 0
(expiration disabled)
The validity period of the failed request counter. Once the configured time has passed since the last failed request to the backend, the counter is reset to zero.
ResourceManager.DisablePing
Type: boolean
Usage Constraints: optional, advanced
Default: false
; and if only one backend is set, and CheckForMultipleIPAddresses
is false
, the default is true
.
This disables ping on startup and on retry.
RequestTimeout
Type: integer [msec]
Usage Constraints: optional, scaling
Default: 120000
(2 min)
The TCP timeout, in milliseconds, for the reading of the response from the web server.
ConnectTimeout
Type: integer [msec]
Usage Constraints: optional, scaling
Default: 15000
(15 sec)
Timeout in milliseconds to open the TCP connection to the content provider.
Lower this timeout for fast recovery (in conjunction with a short ResourceManager.RetryTimeout
).
Raise this timeout if your server has networking problems and the proxy receives "connection refused" errors, even though the server is up, running, and responding to connect requests.
KeepAlive
Type: boolean
Usage Constraints: optional, scaling/tuning
Default: true
Secure default: true
Pool TCP connections for later reuse. If set to false
, the connection is closed after use, and a new TCP connection will be established for the next request.
The web server should support keepalive in an acceptable range. If this is not the case, keepalive should be disabled.
KeepAlive.LifeTime
Type: integer [sec]
Usage Constraints: optional, scaling
Default: 600
The absolute lifetime of a TCP connection if connection pooling is enabled (see KeepAlive
). The connection is closed when:
time(now) - time(connect) >= KeepAliveLifeTime
This parameter is of interest when a firewall between nevisProxy and a content provider drops the TCP connection.
KeepAlive.ConnectionPoolSize
Type: integer
Usage Constraints: optional, scaling, min: 0
, max: 604800
Default: 50
With the config attribute KeepAlive.ConnectionPoolSize
the number of pooled TCP connections can be configured. A TCP connection is only put in the pool if it does not exceed the configured size.
KeepAlive.InactiveInterval
Type: integer
Usage Constraints: optional, scaling
Default: 60
With the KeepAlive.InactiveInterval
, it is possible to establish an upper limit for the HTTP session cache at runtime.
If a backend sends max=0
in a keep-alive
response, the connection will be closed.
If the backend sends its own timeout in a keep-alive
connection, this timeout will overwrite the KeepAlive.InactiveInterval
of the HttpConnectorServlet.
KeepAlive.ByClient
Type: boolean
Usage Constraints: optional, advanced
Default: false
If KeepAlive is set to true
, the TCP connection will only be reused for the same client. A call from a different client will use another TCP connection from the connection pool.
Restriction: TLS caching per session (SSLCache=on
) and connection pooling by client (KeepAlive.ByClient=true
) is not supported.
KeepAlive.RetryOnFail
Type: boolean
Usage Constraints: optional, advanced
Default: true
A risk involved in keep-alive requests is that the backend closes a connection just before the proxy wants to reuse this connection. If you set this attribute to true
, nevisProxy will try to open a new connection in such a situation.
BlockSize
Type: integer [byte]
Usage Constraints: optional, advanced, min: 1
, max: 524288
Default: 8192
The block size used for reading/writing.
(Advanced configuration, as a rule, the default value should be just fine.)
Poll
Type: boolean
Usage Constraints: optional, advanced
Default: false
If this attribute is set to true
, the system forwards the request and response body without caching. This is needed for the Remote Desktop Protocol (RDP) and Outlook Anywhere, which both use RPC over HTTP.
Do not map any buffering filters if the Poll
attribute is set to true
. Otherwise, the error HTTP-0081
will occur. Buffering filters are, for example, rewrite and validation filters.
Protocol
Type: enum
Possible values: HTTP/1.1
Usage Constraints: optional, troubleshooting
Default: not set, browser protocol
This attribute configures the protocol version. If not set, the one on the browser request will be used.
DelegateHeader
Type: string array
Usage Constraints: optional, basic feature
Default: no delegation
This attribute configures additional (constant) HTTP header fields:
<header name>:<header value>
For dynamic header propagation, see the HeaderDelegationFilter.
SuppressHeader
Type: string
Usage Constraints: optional, advanced
Default: no suppression
A space-separated list of (case-insensitive) HTTP header fields that will be suppressed when the request is sent to the content provider.
OutboundProxy
Type: address
Usage Constraints: optional, advanced
A single INET address (<host>:<port>
, where <host>
may be an IP address or a host name) that specifies a forward HTTP proxy to use.
OutboundProxyAuthorization.Username
Type: string
Usage Constraints: optional, advanced
Username for the proxy authorization header.
OutboundProxyAuthorization.Password
Type: string
Usage Constraints: optional, advanced
Password for the proxy authorization header.
DNSCache
Type: boolean
Usage Constraints: optional, troubleshooting
Default: true
Without DNS caching, the DNS name is resolved for every request. You will need to disable this setting if the IP address of your host can change, for example if you do failover switching using DNS.
If you turn off this attribute (DNSCache
) and enable the attribute KeepAlive
, your backend may behave incorrectly if the DNS switches while a connection is being pooled in the HttpConnectorServlet.
For example, the backend may not understand the received session cookies like JSESSION
, because the DNS switched the underlying IP of a given name.
DNSCache.ttl
Type: integer [sec]
Usage Constraints: optional, min: 1
, max: unlimited
Default: 3600
If the attribute DNSCache
is set to true
, this attribute (DNSCache.ttl
) specifies the time period in seconds to cache the DNS info, before the system will retrieve the IP address again.
HostName
Type: string
Usage Constraints: optional, advanced
Default: Hostname of first element in InetAddress
Defines the HTTP header field Host
to send to the content provider. If not set, the host part of the corresponding address in the InetAddress field is used.
You usually need to set this attribute, if you have both a failsafe setup with multiple addresses in the InetAddress field and content rewriting enabled (in the AutoRewrite
attribute).
If the content provider sends a URL with a host part, and the HTTP Host
header is not considered, auto-rewriting does not work properly: Redirects or links will directly point to the content provider instead of to the reverse proxy.
Notes regarding the configuration of the HostName
attribute:
- It is possible to configure several values for this attribute (
\n
-separated). Every value is assigned to the respective value ofInetAddress
. - If this attribute is not configured, the system uses the host part of the
InetAddress
value for the host part of fully qualified URLs. - If the attribute's configuration only contains one value, the system will assign this value to all values of the
InetAddress
attribute. - If failsafeness groups have been defined in the
InetAddress
attribute, theHostName
attribute must contain the same number of groups. - If each failsafeness group defined in the
InetAddress
attribute has the same amount of inet addresses, aHostName
attribute value with the same amount of hosts is valid as well. - Optionally, it is possible to specify the port using the notation
<hostname>:<port>
. - It is possible to define variables from the
ENV
andAUTH
context, for example:ENV:HTTP_Host;
. This makes it possible to use theHost
header sent by the client. - If both the
LogoutURI
parameter and a dynamic hostname are specified (by usingENV
,AUTH
or similar), the system sends the value of theInetAddress
attribute asHost
header on the logout request to the backend.
HostName.StripDefaultPort
Type: boolean
Usage Constraints: optional, advanced, conditions are supported
Supported pragmas: break
Default: true
Defines if the default port should be stripped. If set to false, the Host
header always contains the assigned port number.
HostName.StripNonDefaultPort
Type: boolean
Usage Constraints: optional, advanced, conditions are supported
Supported pragmas: break
Default: false
Defines if ports which are not 80
or 443
should be stripped. If set to false, the Host
header always contains the assigned port number.
ProxyPolicy
Type: enum
Possible values: stealth
, via-header
, mime-completion
, client-side
, commit-response
, commit-response-disabled
Usage Constraints: optional, advanced
Default: stealth
This attribute supports configuration of proxy behavior in a space- or newline-separated list that may contain the following values:
via-header
: Enables setting thevia
response header so that the client can detect whether the HTTP status came from the proxy or from the content provider.stealth
: suppressvia
response header.mime-completion
: If theContent-Type
header of the response is not set, a header with the configured mime-type mapping will be set (workaround for buggy content providers).commit-response
: The response is committed in any case. This prevents the response from being overwritten by a subsequent component (for example an ErrorFilter or similar).commit-response-disabled
: The response is not committed in any case. This ensures that the response can be overwritten by any subsequent component.client-side
: The proxy acts as a client-side proxy for HTTP.
ResponseLineSize
Type: integer [byte]
Usage Constraints: optional, security, scaling, min: 512
, max: 524288
Default: 4096
The maximum allowed line size in the HTTP header of the response.
SetServerHeader
Type: boolean
Usage Constraints: optional, advanced, security
Default: false
If true
, the server
HTTP header of the response will be passed to the client, exposing internal information to the client.
URLEncoding
Type: boolean
Usage Constraints: optional, troubleshooting
Default: true
If set to true
(which is the default), this attribute encodes forbidden characters in the outgoing URI from nevisProxy to the backend application.
For example, if true
, the attribute changes the URI
/UIFont CMSStyle.swift
into
/UIFont%20CMSStyle.swift
This feature can be useful for the integration of content management systems (CMS) that use URLs with forbidden characters.
This attribute is directly related to the low-level attribute org.apache.request.ParsedUri
of the servlet container (see the chapter Low-level properties).
The org.apache.request.ParsedUri
attribute decodes an encoded URI of an incoming request for internal nevisProxy processing.
We highly recommend setting both attributes to true
.
Note: You can set the attribute URLEncoding to false
centrally.
To do this, add the following code lines to the bc.properties configuration file:
ch.nevis.isiweb4.servlet.connector.http.URLEncoding=false
The above code line sets the attribute URLEncoding
to false
for all HttpConnectorServlets and HttpsConnectorServlets, unless it is explicitly overwritten in the servlet itself.
ch.nevis.isiweb4.servlet.connector.http.HttpConnectorServlet.URLEncoding=false
The above code line sets the attribute URLEncoding
to false
for all HttpConnectorServlets, unless it is explicitly overwritten in the servlet itself.
ch.nevis.isiweb4.servlet.connector.http.HttpsConnectorServlet.URLEncoding=false
The above code line sets the attribute URLEncoding
to false
for all HttpsConnectorServlets, unless it is explicitly overwritten in the servlet itself.
ch.nevis.isiweb4.servlet.connector.websocket.URLEncoding=false
The above code line sets the attribute URLEncoding
to false
for all WebsocketServlets, unless it is explicitly overwritten in the servlet itself.
The file bc.properties
is usually located under /var/opt/nevisProxy/<instance name>/conf/bc.properties
.
ResponseRewrite.ParsingMode
Type: enum
Possible values: strict
, tolerant
Usage Constraints: optional, advanced
Default: strict
Sets parsing to either tolerant
or strict
.
If parsing mode is set to strict
, the HTML parser tries to parse improper HTML with heuristics.
It finds any URLs in a text/html
body and (if activated, see below) performs further replacements on any inline JavaScript or CSS parts as if they were inside a text/css
or text/javascript
body.
For the regular expression replacement, the strict
parsing mode makes errors terminate the whole parsing process.
ResponseRewrite.IgnoreInlineCSS
Type: boolean
Usage Constraints: optional, advanced
Default: false
Disables parsing/processing of inline CSS blocks in HTML content.
ResponseRewrite.IgnoreInlineJS
Type: boolean
Usage Constraints: optional, advanced
Default: false
Disables parsing/processing of inline JavaScript blocks in HTML content.
ResponseRewrite.IgnoreHTMLComments
Type: boolean
Usage Constraints: optional, advanced
Default: true
Disables parsing of HTML comments.
ResponseRewrite.PostProcessingURLRegexps
Type: string array
Usage Constraints: optional, advanced
A newline-separated list of regular expressions for the configuration of additional post-processing for URL rewriting.
Works like ResponseRewrite.JavaScriptRegexps
and ResponseRewrite.StyleSheetRegexps
and permits identification of additional URLs in any content defined in ResponseRewrite.PostProcessing.ContentTypes
.
The default regex-type is PCRE(da)
.
ResponseRewrite.PostProcessing.ContentTypes
Type: string
Usage Constraints: optional, advanced
default: ^text/
A newline separated list of regular expressions which define the content types on which the ResponseRewrite.PostProcessingURLRegexps
applies. The default regex-type is PCRE(da)
.
ResponseRewrite.PostProcessingDelimiters
Type: string array
Usage Constraints: optional, advanced
Default: 10,13
(\n \r
)
Supports configuration of alternative delimiters for the tokenization of the response body for ResponseRewrite.PostProcessingURLRegexps
.
ResponseRewrite.PostProcessingURLRegexps
is the only rewriting that supports specification of delimiters, all other ResponseRewrite
options tokenize by line or statement (;
).
The configured regexp (ResponseRewrite.PostProcessingDelimiters
) is executed on the resulting tokens of the body. Delimiters must be configured as a whitespace- or ,
-separated list of decimal ASCII characters.
ResponseRewrite.PostProcessingIgnoreQuotes
Type: boolean
Usage Constraints: optional, advanced
Default: true
Ignores single and double quotes for tokenizing during post-processing.
If quoting is active (ResponseRewrite.PostProcessingIgnoreQuotes=false
), text inside matching quotes is ignored. This includes the other type of quote as well as escaped quotes. Quotes can be escaped with a backslash.
ResponseRewrite.PostProcessingCaseInsensitive
Type: boolean
Usage Constraints: optional, advanced
Default: true
Makes regular expressions in ResponseRewrite.PostProcessingURLRegexps
case-insensitive.
ResponseRewrite.ContentTypes.*
Type: string
Usage Constraints: optional, advanced
A newline-separated list of regular expressions which defines the given ContentType. For example:
ContentTypes.json=^application/json
For each *
at least the ContentType.*.rules
has to be defined.
Defaults are:
^text/html
^application/xhtml
^text/css
^text/javascript
^application/javascript
^application/x-javascript
ResponseRewrite.ContentType.*.rules
Type: string
Usage Constraints: optional, advanced
Defines a list of regular expressions for identifying URLs inside the given content type. The default regex-type is PCRE(da)
.
The system will ignore this parameter if the content type is html
(ResponseRewrite.ContentType.html.rules
).
To enable the evaluation of the rules for content type HTML, set the parameter ResponseRewrite.ContentType.html.rules.enabled
.
ResponseRewrite.ContentType.html.rules.enabled
Type: boolean
Usage Constraints: optional, advanced
Default: false
If this parameter is set to true
, the system will also evaluate the parameter ResponseRewrite.ContentType.*.rules
for content type HTML (ResponseRewrite.ContentType.html.rules
).
When you configure the parameter ResponseRewrite.ContentType.*.rules
for HTML, keep in mind that there are already some built-in HTML rules.
If you add these built-in rules to the configuration of ResponseRewrite.ContentType.html.rules
, the system will rewrite the URLs twice.
These targets are rewritten automatically: HREF
, SRC
, ICONSRC
, ACTION
, BACKGROUND
, CODEBASE
, CODE
, IMAGEPATH
, LOWSRC
, and ARCHIVE
.
ResponseRewrite.ContentType.*.rules.ignoreQuotes
Type: string
Usage Constraints: optional, advanced
Ignores quotes as parenthesis around a string or similar inside the given ContentType code.
ResponseRewrite.ContentType.*.separators
Type: string
Usage Constraints: optional, advanced
Defines the line-separators inside the given ContentType.
ResponseRewrite.MaxMatchLen
Type: integer
Usage Constraints: optional, advanced
Default: 2048
Maximum length of a string a regular expression rule may match.
The amount of memory needed for response rewriting grows with the configured value. For performance reasons, the maximum should be kept as low as possible.
When set to zero, there is no limit to the length of a matched URL. Such a setting is not recommended.
AwaitingResponse
Type: boolean
Usage Constraints: optional, advanced
Default: true
If set to false
, the response of the content provider will not be awaited. Set AwaitingResponse
to false
for "sidecalls" done in rewrite-rules such as ^/(.*)$:/$1:S=NotifierServlet:continue
.
AllowedMethods
Type: string
Usage Constraints: optional, advanced
Default: ALL-HTTP
With this parameter, a comma- or white-space-separated list of all allowed HTTP methods can be defined. You can also specify one of these method groups:
ALL-WEBDAV
: all WEBDAV methods can be addressedALL-HTTP
: all HTTP methods can be addressed
Additionally, you can forbid methods included in ALL-WEBDAV
or ALL-HTTP
, by specifying the method to be forbidden with a -
sign in front of it. For example:
ALL-WEBDAV, ALL-HTTP, -CHECKIN
means that all HTTP and WEBDAV methods are allowed, except CHECKIN
.
With the parameter AllowedMethods
, you configure the HTTP methods allowed by nevisProxy to connect to the backend.
This is in contrast to the allowedMethods
attribute of the web application configuration, with which you configure the HTTP methods allowed by the frontend to connect to nevisProxy.
When you modify the AllowedMethods
parameter described here, you may also have to adapt the allowedMethods
attribute of the web application configuration.
For a description of the allowedMethods
attribute, see chapter Web applications.
Per default the following methods are allowed:
GET, POST, HEAD, DELETE, TRACE, CONNECT, OPTIONS, PUT, PATCH
ALL-WEBDAV
methods are:
MERGE, UNCHECKOUT, MKACTIVITY, PROPPATCH, LOCK, CHECKOUT, SEARCH, COPY, MKCOL, MKWORKSPACE, PROPFIND, UPDATE, REBIND, BASELINE-CONTROL, UNBIND, CHECKIN, VERSION-CONTROL, UNLOCK, LABEL, MOVE, ACL, BIND, REPORT
ALL-HTTP
methods are:
GET, POST, HEAD, DELETE, TRACE, CONNECT, OPTIONS, PUT, PATCH
RewriteBufferSize
Type: integer [bytes]
Usage Constraints: optional, advanced
Ddefault: 16384
The size of the internal buffer (in bytes) for buffering HTML tags. Only relevant if the response is rewritten (see AutoRewrite
).
MergeResponseHeaders
Type: boolean
Usage Constraints: optional, advanced
Default: true
If set to true
, this parameter merges response headers with the same name, as described in RFC 2616.
YAML configuration
Place the path to the yaml configuration in the HttpConnector InetAddress parameter by starting with file:// i.e. file:///var/opt/nevisproxy/default/work/WEB-INF/connector1.yml
The root element must be host-groups containing at least one host-group element. The host-group must at least contain one address element which holds one single address (name or IP). Optionally you can also specify the host name right after the address element with a host-name element, which will be set as the host header. The host-name can also be before a host-group element which then specifies the default host name for that group, which still can be overwritten by a host-name element inside the host-group.
Host groups do a fail-over as soon all specified hosts in that group are not accessible. Inside a host group the hosts are load-balanced.
You also can have custom based elements in the yaml file for your own purposes. Those elements can be anywhere in the yaml configuration and must be prefixed with "custom-" to avoid future name clashes.
host-groups:
- host-name: my.host.de
- host-group:
- address: 192.168.1.3
- address: 192.168.2.3
- address: 234.123.1.10
host-name: special.host.ch
- host-name: other.host.de
- host-group:
- address: 168.13.10.3
- host-group:
- address: 172.16.4.10
- host-name: some.host.org
- custom-id: 1.2.3.4.5.6
AutoRewrite Capabilities
The AutoRewrite parameter provides some standard use cases where you do not need a RewriteFilter. Neverthless it comes with some restrictions and some well-known flaws where you have to correct the outcome of the AutoRewrite with a RewriteFilter.
With pathinfo
AutoRewrite can handle absolute and relative URLs as well as port numbers in the hostnames of the absolute URL. It cannot rewrite a relative path without leading slash index.html
, for this you currently need a RewriteFilter. The AutoRewrite also doesn't support path normalization of /../../index.html
.
Example
For all examples the connector is mapped on /example
.
- The Location header is
/custom/index.html
the AutoRewrite will transform this correctly to/example/custom/index.html
. - The Location header is
custom/index.html
the AutoRewrite will not be rewritten at all but should be/example/custom/index.html
. - The Location header is
/../index.html
the AutoRewrite will transform this to/example/../index.html
, but should be/index.html
(no normalizing).
Cookie path
The cookie path AutoRewriting doesn't rewrite path=/
, but should rewrite it to pathinfo. If a sub path is set, it just replaces the path with the pathinfo but doesn't add the sub path.
Example
For all examples the connector is mapped to /example
Set-Cookie path=/
is not rewritten at all, but should bepath=/example
Set-Cookie path=/any/thing
is rewritten topath=/example
, but should bepath=/example/any/thing
Set-Cookie path=/prefix/custom
with URIPrefix/prefix
should be rewritten topath=/example
, but should bepath=/example/custom
Referrer header
AutoRewrite do not rewrite the Referrer header.
Performance decrease with UrlEncryptionFilter
If the AutoRewrite
parameter is set and auto-rewriting is on
, the HttpConnectorServlet sends the body back in different little chunks. This can slow down the UrlEncryptionFilter.
To avoid this, add a buffering LuaFilter to the filter chain as workaround. The buffering LuaFilter collects all chunks up to a certain size, then sends everything back in bigger chunks. Map the LuaFilter in between the UrlEncryptionFilter and the HttpConnectorServlet. For this case, you can use the LuaFilter_buffering.example
configuration, which you can find in the installed nevisProxy package in /opt/nevisproxy/examples/various/LuaFilter_buffering.example
.
NoSessionCookie
The NoSessionCookie feature of the Http(s)ConnectorServlet allows you to implement sticky load balancing without creating real sessions in nevisProxy. Instead of storing the information in the session, an encrypted cookie is sent to the client. This cookie contains all state details of the given Http(s)ConnectorServlet. This is possible because the HttpConnectorState is externalized, encrypted, then signed, and finally added as a response cookie header. When the client sends back the NoSessionCookie with the next request, the Http(s)ConnectorServlet verifies, decrypts and reads the cookie, thus being able to retrieve such state details as the connected backend or the notification URI:
You configure the behavior of the NoSessionCookie feature with the following ResourceManager.NoSessionCookie.*
attributes of the HttpConnectorServlet:
ResourceManager.NoSessionCookie
Type: boolean
Usage Constraints: optional, advanced
Default: false
The information which backend has been used will not be saved within the HTTP session but within a cookie sent to the client. This facilitates sticky load balancing without sessions within nevisProxy.
This parameter cannot be set to true
when SSLCache is set to session
, and either DisableBindingStatusCode
or KeepAlive.ByClient
is enabled.
ResourceManager.NoSessionCookie.CookieName
Type: string
Usage Constraints: optional, advanced
Default: IW4NoSession
The name of the cookie if the parameter ResourceManager.NoSessionCookie
is enabled.
ResourceManager.NoSessionCookie.CookieDomain
Type: string
Usage Constraints: optional, advanced
The domain of the cookie if the parameter ResourceManager.NoSessionCookie
is enabled.
ResourceManager.NoSessionCookie.CookiePath
Type: string
Usage Constraints: optional, advanced
The path of the cookie if the parameter ResourceManager.NoSessionCookie
is enabled.
ResourceManager.NoSessionCookie.CookieHttpOnly
Type: boolean
Usage Constraints: optional, advanced
Default: true
If set to true
, the HttpOnly
attribute is added to the cookie, if the parameter ResourceManager.NoSessionCookie
is enabled.
ResourceManager.NoSessionCookie.CookieSecure
Type: boolean
Usage Constraints: optional, advanced
Default: false
Secure default: true
If set to true
, the ResourceManager.NoSessionCookie
will have the secure attribute set.
ResourceManager.NoSessionCookie.EncryptionKey
Type: string
Usage Constraints: optional, advanced
Symmetric encryption key used to encrypt the information within the cookie (the stickiness information).