This research represents the second version of the Persian Discourse Treebank (PerDTB). The first version was based on nearly 30,000 individual sentences and so imposed some restrictions: the NoRel relation type in the Penn Discourse Treebank (PDTB) schema was not defined in this corpus; more than half of the sentences were simple clauses containing minimal discourse information and, more importantly, due to the use of individual sentences, the investigation of intra-sentential relations was not actually possible. So, in the second version, we collected 1,383 documents with more than half a million words, mainly crawled from news articles. They were manually annotated with morphological and syntactic labels, and we then performed discourse analysis based on the PDTB schema. Also, to enrich our data, we added NER and coreference information at the top. In discourse analysis, we explored any semantic or logical relation between sentences (inter-sentential) or between clauses inside a complex sentence (intra-sentential), resulting in the annotation of 24,502 discourse relations along with their senses, arguments, and their attributes, including type, source, scopal polarity, determinacy, and modality. To measure inter-annotator agreement, we double-annotated 229 documents with 107,653 tokens. Inter-annotator agreement was good for sense classification (κ = 0.753) and very good for Implicit, EntRel, and Clause relations (κ = 0.809).